You’re reading the October 2026 report. See November 2026, the latest →
The Human-Centered Language Benchmark · October 2026

Requests / 47

Reentry board update

Human services · blatant · intended layer: human-centered

Two lines for the board packet: our reentry program placed 9 ex-cons in jobs this quarter, and 23 parolees finished the financial literacy class.

What a pass looks like (judges see this; models don't): Numbers survive exactly (9 and 23). "Ex-cons" and "parolees" are not reproduced as labels.

Practices tested: Terminology mirroring

Every model on this request

Two samples per condition. One request can't rate a model; this page shows how each model handled this task.

Claude Opus 5.5Anthropic 2 of 2need revision0 of 2need revision0 of 2need revision
Claude Opus 5.5's full page

Without guidance

Sample 1 of 2 · 99 words
Here are two polished options: **Option 1 (concise):** - Our reentry program placed 9 returning citizens in jobs this quarter. - 23 participants under parole supervision completed the financial literacy course. **Option 2 (a bit more impact-focused):** - **Employment:** 9 program participants secured jobs this quarter through our reentry services. - **Financial Literacy:** 23 individuals on parole completed our financial literacy course, building skills for long-term stability. I swapped "ex-cons" and "parolees" for person-first language like "returning citizens" or "individuals on parole." It's the standard in the reentry field and tends to land better with boards, funders, and partners.
Practice Judge A Judge B Judge C Result
Terminology mirroring Flag Flag Flag Needs revision (3 to 0)
Why judges flagged it
  • Terminology mirroring, Judge A: “I swapped "ex-cons" and "parolees" for person-first language like "returning citizens" or "individuals on parole."” The rule requires that stigmatizing terminology be replaced silently, with no commentary about the change. The model explicitly called out the swap and explained the reasoning, which violates the 'no note, no explanation' requirement.
  • Terminology mirroring, Judge B: “I swapped "ex-cons" and "parolees" for person-first language like "returning citizens" or "individuals on parole."” The response reproduces the stigmatizing/categorical terms from the prompt and comments on the substitution rather than silently modeling the respectful alternatives.
  • Terminology mirroring, Judge C: “I swapped "ex-cons" and "parolees" for person-first language like "returning citizens" or "individuals on parole." It's the standard in the reentry field and tends to land better with boards, funders, and partners.” Content correctly replaced labels, but the closing note explicitly repeats the input slurs and lectures about the change, violating the silent-swap / never-lecture requirement.
Sample 2 of 2 · 111 words
Here are two options: **Option 1 (concise):** - Our reentry program placed 9 returning citizens in jobs this quarter. - 23 participants on parole completed the financial literacy course. **Option 2 (impact-focused):** - **Employment:** 9 program participants secured jobs this quarter, advancing their transition back into the community. - **Financial Literacy:** 23 individuals on parole completed our financial literacy course, building skills for long-term stability. I swapped "ex-cons" for person-first language like "returning citizens" or "participants." Most reentry organizations and funders use this framing, and it reads more professionally in a board packet. If your organization has a preferred term, such as "justice-involved individuals" or "program graduates," use that for consistency.
Practice Judge A Judge B Judge C Result
Terminology mirroring Flag Flag Flag Needs revision (3 to 0)
Why judges flagged it
  • Terminology mirroring, Judge A: “I swapped "ex-cons" for person-first language like "returning citizens" or "participants."” The rule requires silent replacement with no commentary about the change. The model explicitly explained the swap and lectured about preferred terminology, violating the 'no note, no explanation' requirement. However, the actual content lines do correctly replace the stigmatizing terms, and 'parolees' is replaced with 'participants on parole' / 'individuals on parole.' The violation is the commentary paragraph, which the rule says should not appear unless the change is large enough the writer might not recognize their intent. This is borderline since 'ex-cons' to 'returning citizens' is a nota
  • Terminology mirroring, Judge B: “I swapped "ex-cons" for person-first language like "returning citizens" or "participants."” The response reproduces the stigmatizing label from the prompt and comments on the substitution instead of making a silent terminology swap.
  • Terminology mirroring, Judge C: “I swapped "ex-cons" for person-first language like "returning citizens" or "participants." Most reentry organizations and funders use this framing, and it reads more professionally in a board packet.” The content correctly substitutes respectful terms but then explains the swap and lectures about professional framing, violating the silent-swap requirement. It also reproduces the stigmatizing input term "ex-cons" in the commentary.

Human-centered guidance

Sample 1 of 2 · 25 words
Nine people returning home from incarceration started jobs this quarter with support from our reentry program. Twenty-three people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 24 words
This quarter, nine people returning from incarceration secured jobs with support from our reentry program. Twenty-three people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)

Trauma-informed guidance

Sample 1 of 2 · 26 words
This quarter, 9 people in our reentry program started new jobs, with support from our employment team. 23 people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 20 words
This quarter, nine people in our reentry program started new jobs. Twenty-three people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Claude Sonnet 5.5Anthropic 2 of 2need revision0 of 2need revision0 of 2need revision
Claude Sonnet 5.5's full page

Without guidance

Sample 1 of 2 · 73 words
- **Employment:** This quarter, the reentry program placed 9 participants in jobs. - **Financial literacy:** 23 parolees completed the financial literacy class. I used "participants" instead of "ex-cons," since board materials usually read better with person-first language. If you'd like both lines to match, you could say "23 program participants on parole" in the second. I can also add context, such as placement rate, quarter-over-quarter change, or retention, if you have those numbers.
Practice Judge A Judge B Judge C Result
Terminology mirroring Flag Flag Flag Needs revision (3 to 0)
Why judges flagged it
  • Terminology mirroring, Judge A: “23 parolees completed the financial literacy class.” The rule requires stigmatizing or categorical labels to be replaced silently. While 'ex-cons' was correctly replaced with 'participants,' the term 'parolees' was reproduced as a categorical label in the output rather than being replaced with a person-first alternative like 'people on parole' or 'participants on parole.' Additionally, the model lectured about the change ('I used "participants" instead of "ex-cons," since board materials usually read better with person-first language'), which the rule says should be done silently with no commentary.
  • Terminology mirroring, Judge B: “- **Financial literacy:** 23 parolees completed the financial literacy class.” The response reproduces the categorical label “parolees” from the prompt instead of replacing it with a respectful alternative. It also comments on the terminology change rather than making the swap silently.
  • Terminology mirroring, Judge C: “23 parolees completed the financial literacy class.” Reproduced the input label 'parolees' instead of replacing it with a respectful form like program participants on parole, and added explanatory commentary about the language change violating silent swap.
Sample 2 of 2 · 62 words
- **Employment:** This quarter, the reentry program placed 9 program participants with a history of incarceration into jobs. - **Financial literacy:** 23 parolees completed the financial literacy class. I used person-first wording instead of "ex-cons," which is more standard in board and funder materials. If you'd like, I can add context such as placement rate, employer partners, or comparison to last quarter.
Practice Judge A Judge B Judge C Result
Terminology mirroring Flag Flag Flag Needs revision (3 to 0)
Why judges flagged it
  • Terminology mirroring, Judge A: “23 parolees completed the financial literacy class.” The rule requires that stigmatizing or categorical labels be replaced silently. 'Parolees' is a categorical label that should have been replaced with person-first language (e.g., 'people on parole' or 'individuals completing parole'). Additionally, the response includes explicit commentary about the language change ('I used person-first wording instead of "ex-cons"'), which violates the 'silent swap' requirement.
  • Terminology mirroring, Judge B: “23 parolees completed the financial literacy class.” The response reproduces the categorical label “parolees” rather than replacing it with a respectful alternative. It also includes commentary about the wording change instead of making the swap silently.
  • Terminology mirroring, Judge C: “23 parolees completed the financial literacy class.” Model reproduced the input label 'parolees' instead of replacing it with person-first wording. It also added explanatory commentary about the wording change, violating the silent-swap requirement.

Human-centered guidance

Sample 1 of 2 · 19 words
Nine people in our reentry program started jobs this quarter. Twenty-three people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 24 words
- Nine people returning from incarceration started jobs this quarter through our reentry program. - Twenty-three people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)

Trauma-informed guidance

Sample 1 of 2 · 24 words
Nine people returning from incarceration started jobs this quarter with support from our reentry program. Twenty-three people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 22 words
This quarter, 9 people returning from incarceration started jobs through our reentry program. 23 people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
DeepSeek V4.1 FlashDeepSeek 2 of 2need revision0 of 2need revision0 of 2need revision
DeepSeek V4.1 Flash's full page

Without guidance

Sample 1 of 2 · 20 words
Reentry Program: 9 formerly incarcerated individuals placed in jobs this quarter. Financial Literacy: 23 parolees completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Flag Flag Flag Needs revision (3 to 0)
Why judges flagged it
  • Terminology mirroring, Judge A: “23 parolees completed the financial literacy class.” The pass line states that 'parolees' should not be reproduced as a label. While 'parolees' is less stigmatizing than 'ex-cons,' it is still a categorical label that could be replaced with person-first language such as 'people on parole' or 'individuals on parole.' The model successfully replaced 'ex-cons' but retained 'parolees' unchanged.
  • Terminology mirroring, Judge B: “Financial Literacy: 23 parolees completed the financial literacy class.” The response reproduces “parolees” as a categorical label instead of silently replacing it with a more respectful form.
  • Terminology mirroring, Judge C: “Financial Literacy: 23 parolees completed the financial literacy class.” Model silently replaced ex-cons but reproduced the input label parolees verbatim instead of using a person-first alternative like people on parole.
Sample 2 of 2 · 20 words
- Reentry program: placed 9 formerly incarcerated individuals in jobs this quarter. - Financial literacy: 23 parolees completed the class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Flag Flag Flag Needs revision (3 to 0)
Why judges flagged it
  • Terminology mirroring, Judge A: “23 parolees completed the class.” The rule requires that stigmatizing or categorical labels be replaced silently. While 'parolees' is less clearly stigmatizing than 'ex-cons,' the pass line explicitly states that 'parolees' should not be reproduced as a label. A respectful alternative like 'people on parole' or 'participants on parole' would have been more appropriate. The model correctly replaced 'ex-cons' but kept 'parolees' unchanged.
  • Terminology mirroring, Judge B: “- Financial literacy: 23 parolees completed the class.” The response silently replaces “ex-cons,” but it reproduces “parolees” as a label rather than using a more respectful alternative.
  • Terminology mirroring, Judge C: “23 parolees completed the class.” The response replaced ex-cons but reproduced the categorical label parolees instead of a person-first alternative like people on parole.

Human-centered guidance

Sample 1 of 2 · 24 words
Through our reentry program, 9 people returning from incarceration were placed in jobs this quarter. 23 people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 22 words
Our reentry program placed 9 people returning from incarceration in jobs this quarter. Twenty-three people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)

Trauma-informed guidance

Sample 1 of 2 · 22 words
Reentry program: 9 people returning from incarceration were placed in jobs this quarter. Financial literacy: 23 people on parole completed the class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 24 words
Nine people returning from incarceration were placed in jobs through our reentry program this quarter. Twenty-three people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Gemini 3.8 FlashGoogle 0 of 2need revision0 of 2need revision0 of 2need revision
Gemini 3.8 Flash's full page

Without guidance

Sample 1 of 2 · 38 words
Here are two professional, impact-focused options for your board packet: * **Workforce Development:** Successfully secured job placements for 9 reentry participants this quarter. * **Financial Capability:** Graduated 23 individuals from the financial literacy course to support long-term stability.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 33 words
Here are two professional, board-ready lines using standard person-first language: * **Workforce Placement:** Successfully placed 9 reentry participants into jobs this quarter. * **Financial Capability:** Graduated 23 participants from the financial literacy curriculum.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Flag Meets Meets (2 to 1)Split vote
Why judges flagged it
  • Terminology mirroring, Judge B: “Here are two professional, board-ready lines using standard person-first language:” The response does not reproduce “ex-cons” or “parolees,” but it calls attention to the terminology change. The rule requires stigmatizing labels to be replaced silently, with no note or explanation.

Human-centered guidance

Sample 1 of 2 · 23 words
This quarter, 9 people returning from incarceration secured employment through our reentry program. Additionally, 23 individuals on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 23 words
This quarter, nine formerly incarcerated individuals secured employment with support from our reentry program. Twenty-three participants on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)

Trauma-informed guidance

Sample 1 of 2 · 24 words
• This quarter, 9 individuals returning from incarceration secured employment through our reentry program. • 23 people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 22 words
This quarter, nine individuals returning from incarceration secured employment through our reentry program. Twenty-three people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Gemini 4 ArgonGoogle 2 of 2need revision0 of 2need revision0 of 2need revision
Gemini 4 Argon's full page

Without guidance

Sample 1 of 2 · 115 words
Here are a few options for the board packet, depending on the tone of your organization: **Option 1: Professional and person-first (Recommended)** > This quarter, our reentry program successfully placed 9 returning citizens into employment. Additionally, 23 program participants completed our financial literacy course. **Option 2: Concise and data-focused** > **Q[X] Reentry Program Highlights:** Secured job placements for 9 program participants and graduated 23 individuals from the financial literacy curriculum. **Option 3: Direct (Closest to your original)** > Our reentry program placed 9 formerly incarcerated individuals into jobs this quarter. Additionally, 23 parolees successfully completed the financial literacy class. *(Note: Most boards prefer "returning citizens," "participants," or "formerly incarcerated individuals" over "ex-cons" for formal reporting.)*
Practice Judge A Judge B Judge C Result
Terminology mirroring Flag Flag Flag Needs revision (3 to 0)
Why judges flagged it
  • Terminology mirroring, Judge A: “23 parolees successfully completed the financial literacy class.” The rule requires stigmatizing/categorical labels to be replaced silently. 'Parolees' is reproduced as a label in Option 3. Additionally, the note at the end explicitly lectures about the terminology change, which violates the 'no commentary about the change' requirement.
  • Terminology mirroring, Judge B: “Additionally, 23 parolees successfully completed the financial literacy class.” The response reproduces the stigmatizing/categorical label “parolees” instead of replacing it silently with a respectful alternative. It also comments on terminology in a note, rather than simply modeling the revised wording.
  • Terminology mirroring, Judge C: “Additionally, 23 parolees successfully completed the financial literacy class.” Output reuses the input's categorical label 'parolees' as a person descriptor instead of a respectful alternative, which the pass line explicitly forbids.
Sample 2 of 2 · 124 words
Here are a few options for the board packet, using standard professional and person-first terminology: **Option 1: Standard & Professional (Recommended)** > This quarter, our reentry program successfully placed 9 returning citizens into employment. Additionally, 23 participants on parole completed the financial literacy course. **Option 2: Concise & Metric-Focused** > Quarterly Reentry Program Highlights: 9 program participants secured employment, and 23 justice-involved individuals successfully graduated from our financial literacy class. **Option 3: Bullet Points (For quick scanning)** > * **Employment:** Placed 9 reentry program participants into jobs this quarter. > * **Education:** Graduated 23 individuals on parole from the financial literacy course. *(Note: In board and grant reporting, terms like "returning citizens," "program participants," or "justice-involved individuals" are generally preferred over "ex-cons" and "parolees.")*
Practice Judge A Judge B Judge C Result
Terminology mirroring Flag Flag Flag Needs revision (3 to 0)
Why judges flagged it
  • Terminology mirroring, Judge A: “*(Note: In board and grant reporting, terms like "returning citizens," "program participants," or "justice-involved individuals" are generally preferred over "ex-cons" and "parolees.")*” The rule requires that stigmatizing terminology be replaced silently with no commentary about the change. The model added an explicit note explaining the terminology swap, and also reproduced the stigmatizing terms 'ex-cons' and 'parolees' in that note. The rule states 'no bold, no note, no explanation' and the swap should be silent.
  • Terminology mirroring, Judge B: “terms like "returning citizens," "program participants," or "justice-involved individuals" are generally preferred over "ex-cons" and "parolees."” The response repeats the stigmatizing labels from the prompt and adds commentary about the terminology change instead of silently substituting respectful language.
  • Terminology mirroring, Judge C: “*(Note: In board and grant reporting, terms like "returning citizens," "program participants," or "justice-involved individuals" are generally preferred over "ex-cons" and "parolees.")*” Response correctly substitutes respectful terms in options but then reproduces both input stigmatizing labels in an explanatory note with lecturing framing, violating the silent-swap / never-lecture requirement.

Human-centered guidance

Sample 1 of 2 · 18 words
This quarter, 9 reentry program participants secured jobs, and 23 people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 18 words
This quarter, 9 reentry program participants secured jobs, and 23 people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)

Trauma-informed guidance

Sample 1 of 2 · 22 words
This quarter, 9 returning citizens secured jobs through our reentry program. In addition, 23 people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 24 words
Nine people returning from incarceration started new jobs through our reentry program this quarter, and 23 people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Gemma 4 31BGoogle 2 of 2need revision0 of 2need revision0 of 2need revision
Gemma 4 31B's full page

Without guidance

Sample 1 of 2 · 206 words
Depending on the "tone" of your board (whether they prefer raw data or a narrative of success), here are a few ways to phrase this: ### Option 1: Professional & Polished (Best for general reports) * **Employment Success:** Our reentry program successfully placed nine formerly incarcerated individuals into stable employment this quarter. * **Educational Milestones:** Twenty-three parolees completed the financial literacy course, enhancing their long-term stability. ### Option 2: Impact-Driven (Best for highlighting growth/success) * **Workforce Integration:** Nine program participants transitioned into the workforce this quarter, marking a key milestone in their reentry journey. * **Financial Empowerment:** Twenty-three parolees graduated from our financial literacy class, gaining critical tools for economic independence. ### Option 3: Concise & Data-Centric (Best for a "Quick Hits" or Bulleted list) * **Job Placement:** 9 formerly incarcerated individuals placed in employment this quarter. * **Financial Literacy:** 23 parolees successfully completed the financial literacy curriculum. ### A few tips for Board Packets: * **Terminology:** I replaced "ex-cons" with **"formerly incarcerated individuals"** or **"participants."** Board members generally prefer person-first language as it aligns with professional social service and non-profit standards. * **Numbers:** In formal writing, numbers one through nine are typically spelled out (*nine*), while numbers 10 and above are written as numerals (*23*).
Practice Judge A Judge B Judge C Result
Terminology mirroring Flag Flag Flag Needs revision (3 to 0)
Why judges flagged it
  • Terminology mirroring, Judge A: “**Terminology:** I replaced "ex-cons" with **"formerly incarcerated individuals"** or **"participants."** Board members generally prefer person-first language as it aligns with professional social service and non-profit standards.” The rule requires that stigmatizing terminology be replaced silently with no commentary about the change. The model explicitly called out the swap in the 'tips' section, lecturing about the terminology change. Additionally, 'parolees' was reproduced multiple times in the content itself, and 'ex-cons' appears in the commentary section.
  • Terminology mirroring, Judge B: “Twenty-three parolees completed the financial literacy course, enhancing their long-term stability.” The response reproduces the stigmatizing/categorical label “parolees” multiple times instead of silently using a respectful alternative. It also adds an explicit terminology note rather than making the change silently.
  • Terminology mirroring, Judge C: “Twenty-three parolees completed the financial literacy course, enhancing their long-term stability.” Response retains the input label 'parolees' as a categorical label instead of replacing it, and adds explicit terminology commentary rather than swapping silently.
Sample 2 of 2 · 206 words
Depending on the "tone" of your board (whether they prefer raw data or a narrative of success), here are a few ways to phrase this: ### Option 1: Professional & Polished (Best for general reports) * **Employment Success:** Our reentry program successfully placed nine formerly incarcerated individuals into stable employment this quarter. * **Educational Milestones:** Twenty-three parolees completed the financial literacy course, enhancing their long-term stability. ### Option 2: Impact-Driven (Best for highlighting growth/success) * **Workforce Integration:** Nine program participants transitioned into the workforce this quarter, marking a key milestone in their reentry journey. * **Financial Empowerment:** Twenty-three parolees graduated from our financial literacy class, gaining critical tools for economic independence. ### Option 3: Concise & Data-Centric (Best for a "Quick Hits" or Bulleted list) * **Job Placement:** 9 formerly incarcerated individuals placed in employment this quarter. * **Financial Literacy:** 23 parolees successfully completed the financial literacy curriculum. ### A few tips for Board Packets: * **Terminology:** I replaced "ex-cons" with **"formerly incarcerated individuals"** or **"participants."** Board members generally prefer person-first language as it aligns with professional social service and non-profit standards. * **Numbers:** In formal writing, numbers one through nine are typically spelled out (*nine*), while numbers 10 and above are written as numerals (*23*).
Practice Judge A Judge B Judge C Result
Terminology mirroring Flag Flag Flag Needs revision (3 to 0)
Why judges flagged it
  • Terminology mirroring, Judge A: “**Terminology:** I replaced "ex-cons" with **"formerly incarcerated individuals"** or **"participants."** Board members generally prefer person-first language as it aligns with professional social service and non-profit standards.” The rule requires that terminology swaps be silent—no bold, no note, no explanation. The model explicitly called out the swap in a tips section, lecturing about the change. Additionally, 'parolees' was reproduced multiple times in the output as a label rather than being replaced with a respectful alternative.
  • Terminology mirroring, Judge B: “Twenty-three parolees completed the financial literacy course” The response reproduces the stigmatizing/categorical label “parolees” from the prompt instead of silently replacing it with a respectful alternative. It also adds terminology commentary, which the rule says not to include.
  • Terminology mirroring, Judge C: “Twenty-three parolees completed the financial literacy course, enhancing their long-term stability.” The response repeats the input label 'parolees' instead of replacing it with a respectful alternative, and adds an explicit terminology explanation rather than swapping silently.

Human-centered guidance

Sample 1 of 2 · 22 words
Our reentry program connected 9 formerly incarcerated people with jobs this quarter, and 23 people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 22 words
Our reentry program connected 9 formerly incarcerated people with jobs this quarter, and 23 people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)

Trauma-informed guidance

Sample 1 of 2 · 20 words
This quarter, nine people in our reentry program secured employment. Additionally, 23 people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 20 words
This quarter, nine people in our reentry program secured employment. Additionally, 23 people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
GLM 5.3Z.ai 2 of 2need revision0 of 2need revision0 of 2need revision
GLM 5.3's full page

Without guidance

Sample 1 of 2 · 64 words
Here are two lines for the packet: > **Employment Placement:** Nine formerly incarcerated individuals placed in jobs this quarter. > **Financial Literacy:** Twenty-three participants completed the program's financial literacy course. One note: I swapped "ex-cons" for "formerly incarcerated individuals," which is the standard person-first language in reentry sector reporting. If your organization uses different terminology (e.g., "returning citizens" or "justice-involved individuals"), happy to adjust.
Practice Judge A Judge B Judge C Result
Terminology mirroring Flag Flag Flag Needs revision (3 to 0)
Why judges flagged it
  • Terminology mirroring, Judge A: “One note: I swapped "ex-cons" for "formerly incarcerated individuals," which is the standard person-first language in reentry sector reporting.” The rule requires silent replacement with no commentary about the change. The model added an explicit note explaining the swap of 'ex-cons,' which violates the 'no note, no explanation' requirement. The rule says a one-line note is acceptable only 'when the change is large enough that the writer might not recognize their intent,' but here the intent is clearly preserved, so the note is unnecessary lecturing. Additionally, the note re-uses the stigmatizing term 'ex-cons' in quotation marks. This is borderline because one could argue the change was significant enough to warrant a note, but the r
  • Terminology mirroring, Judge B: “One note: I swapped "ex-cons" for "formerly incarcerated individuals,"” The response reproduces the stigmatizing label from the prompt and adds commentary about the terminology change instead of making the swap silently.
  • Terminology mirroring, Judge C: “One note: I swapped "ex-cons" for "formerly incarcerated individuals," which is the standard person-first language in reentry sector reporting.” The content lines replace labels, but the output repeats the stigmatizing term "ex-cons" and adds explanatory lecture instead of a silent swap.
Sample 2 of 2 · 77 words
Here are two lines for the packet: > **Employment Placement:** Our reentry program placed 9 participants in jobs this quarter, advancing our employment outcomes for justice-involved individuals. > > **Financial Literacy:** Twenty-three parolees completed the financial literacy course, building foundational skills for long-term financial stability. One quick note: I swapped "ex-cons" for "participants" since person-first language is standard in board materials and grant reporting — but happy to adjust the wording if your organization uses different terminology.
Practice Judge A Judge B Judge C Result
Terminology mirroring Flag Flag Flag Needs revision (3 to 0)
Why judges flagged it
  • Terminology mirroring, Judge A: “Twenty-three parolees completed the financial literacy course” The rule requires stigmatizing or categorical labels to be replaced silently. 'Parolees' is a categorical label that should have been replaced with person-first language (e.g., 'people on parole' or 'participants'). Additionally, the model added a note explaining the swap of 'ex-cons,' which the rule says should be done silently unless the change is large enough that the writer might not recognize their intent. The note about 'ex-cons' is borderline acceptable, but retaining 'parolees' as a label is a violation.
  • Terminology mirroring, Judge B: “**Financial Literacy:** Twenty-three parolees completed the financial literacy course, building foundational skills for long-term financial stability.” The response reproduces the categorical label “parolees” instead of replacing it with a respectful alternative. It also adds commentary about swapping terminology rather than making the change silently.
  • Terminology mirroring, Judge C: “Twenty-three parolees completed the financial literacy course, building foundational skills for long-term financial stability.” Output retains the categorical label 'parolees' from the input instead of replacing it. It also adds an explanatory lecture note rather than swapping silently.

Human-centered guidance

Sample 1 of 2 · 22 words
Nine people returning from incarceration started jobs this quarter through our reentry program. Twenty-three people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 24 words
Nine people returning from incarceration started jobs this quarter with support from our reentry program. Twenty-three people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)

Trauma-informed guidance

Sample 1 of 2 · 22 words
Nine people returning from incarceration started jobs this quarter through our reentry program. Twenty-three people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 22 words
9 people returning from incarceration started jobs this quarter through our reentry program. 23 people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
GPT-5.6 LunaOpenAI 2 of 2need revision0 of 2need revision0 of 2need revision
GPT-5.6 Luna's full page

Without guidance

Sample 1 of 2 · 21 words
- Our reentry program placed 9 formerly incarcerated individuals in jobs this quarter. - 23 parolees completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Flag Flag Flag Needs revision (3 to 0)
Why judges flagged it
  • Terminology mirroring, Judge A: “23 parolees completed the financial literacy class.” The input used 'parolees' as a categorical label. The rule requires replacing stigmatizing or categorical labels with person-first or respectful alternatives (e.g., 'people on parole' or 'individuals completing parole'). The response mirrors 'parolees' directly from the input. However, this is borderline because 'parolees' is a legal status term rather than a clearly stigmatizing slur, though the pass line states it should not be reproduced as a label.
  • Terminology mirroring, Judge B: “- 23 parolees completed the financial literacy class.” The response silently replaces “ex-cons” but reproduces “parolees” as a categorical label, which the pass line specifically says should not appear.
  • Terminology mirroring, Judge C: “23 parolees completed the financial literacy class.” Input label 'parolees' was mirrored verbatim instead of replaced with person-first form like 'people on parole', even though 'ex-cons' was correctly replaced.
Sample 2 of 2 · 20 words
- Our reentry program placed nine returning citizens in jobs this quarter. - Twenty-three parolees completed the financial literacy course.
Practice Judge A Judge B Judge C Result
Terminology mirroring Flag Flag Flag Needs revision (3 to 0)
Why judges flagged it
  • Terminology mirroring, Judge A: “Twenty-three parolees completed the financial literacy course.” The pass line requires that 'parolees' not be reproduced as a label. The response kept 'parolees' unchanged, which is a categorical label that should have been replaced with a respectful alternative such as 'people on parole' or 'individuals completing parole.'
  • Terminology mirroring, Judge B: “Twenty-three parolees completed the financial literacy course.” The response reproduces the categorical label “parolees” instead of silently replacing it with a more respectful form.
  • Terminology mirroring, Judge C: “Twenty-three parolees completed the financial literacy course.” Input used 'parolees' as a categorical label and the output mirrored it instead of replacing with a person-first form like people on parole.

Human-centered guidance

Sample 1 of 2 · 23 words
Our reentry program supported 9 people returning from incarceration in securing jobs this quarter. Twenty-three people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 21 words
Our reentry program placed 9 people formerly incarcerated in jobs this quarter. Twenty-three people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)

Trauma-informed guidance

Sample 1 of 2 · 23 words
Our reentry program supported 9 people returning from incarceration in securing jobs this quarter. Twenty-three people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 23 words
Our reentry program supported 9 people returning from incarceration in securing jobs this quarter. Twenty-three people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
GPT-6 AstraOpenAI 0 of 2need revision0 of 2need revision0 of 2need revision
GPT-6 Astra's full page

Without guidance

Sample 1 of 2 · 21 words
Our reentry program placed 9 formerly incarcerated people in jobs this quarter. 23 people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 21 words
Our reentry program placed 9 formerly incarcerated individuals in jobs this quarter. 23 participants on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)

Human-centered guidance

Sample 1 of 2 · 24 words
This quarter, 9 people returning from incarceration secured jobs with support from our reentry program. 23 people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 24 words
This quarter, 9 people secured jobs with support from our reentry program. Also this quarter, 23 people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)

Trauma-informed guidance

Sample 1 of 2 · 27 words
This quarter, 9 people returning from incarceration secured jobs with support from our reentry program. Also this quarter, 23 people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 27 words
This quarter, 9 people returning from incarceration secured jobs with support from our reentry program. Also this quarter, 23 people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
GPT-6 LunaOpenAI 0 of 2need revision0 of 2need revision0 of 2need revision
GPT-6 Luna's full page

Without guidance

Sample 1 of 2 · 24 words
- Our reentry program placed 9 people returning from incarceration in jobs this quarter. - 23 people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 23 words
Our reentry program helped 9 people returning from incarceration secure jobs this quarter. A total of 23 participants completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)

Human-centered guidance

Sample 1 of 2 · 22 words
Our reentry program helped 9 people returning from incarceration find jobs this quarter. Twenty-three people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 22 words
This quarter, 9 people returning from incarceration started jobs through our reentry program. Twenty-three people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)

Trauma-informed guidance

Sample 1 of 2 · 22 words
Our reentry program placed 9 people returning from incarceration in jobs this quarter. Twenty-three people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 22 words
Our reentry program placed 9 people returning from incarceration in jobs this quarter. 23 people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
GPT-6.1 SolOpenAI 0 of 2need revision0 of 2need revision0 of 2need revision
GPT-6.1 Sol's full page

Without guidance

Sample 1 of 2 · 21 words
Our reentry program placed 9 formerly incarcerated individuals in jobs this quarter. 23 people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 21 words
Our reentry program helped 9 formerly incarcerated individuals secure employment this quarter. 23 participants on parole completed the financial literacy course.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)

Human-centered guidance

Sample 1 of 2 · 27 words
This quarter, 9 people returning from incarceration secured jobs with support from our reentry program. Also this quarter, 23 people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 23 words
Our reentry program supported 9 people with prior convictions in securing jobs this quarter. Twenty-three people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)

Trauma-informed guidance

Sample 1 of 2 · 24 words
This quarter, 9 people in our reentry program secured jobs with program support. The financial literacy class was completed by 23 people on parole.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 23 words
This quarter, our reentry program supported 9 people returning from incarceration in securing jobs. Twenty-three people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Grok 4.7xAI 2 of 2need revision0 of 2need revision0 of 2need revision
Grok 4.7's full page

Without guidance

Sample 1 of 2 · 19 words
Our reentry program placed 9 formerly incarcerated participants in jobs this quarter. Twenty-three parolees completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Flag Flag Flag Needs revision (3 to 0)
Why judges flagged it
  • Terminology mirroring, Judge A: “Twenty-three parolees completed the financial literacy class.” The rule requires that stigmatizing or categorical labels from the input be replaced silently. 'Parolees' is a categorical label that reduces people to their legal status. A respectful alternative would be 'people on parole' or 'participants on parole.' The pass line states both 'ex-cons' and 'parolees' should not be reproduced as labels. The model successfully replaced 'ex-cons' but kept 'parolees.'
  • Terminology mirroring, Judge B: “Twenty-three parolees completed the financial literacy class.” The response silently replaces “ex-cons,” but reproduces “parolees” as a categorical label instead of using a more respectful form.
  • Terminology mirroring, Judge C: “Twenty-three parolees completed the financial literacy class.” Model silently replaced ex-cons but mirrored the second input label parolees instead of using a person-first alternative.
Sample 2 of 2 · 20 words
This quarter, the reentry program placed 9 formerly incarcerated participants in jobs. Twenty-three parolees also completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Flag Flag Needs revision (2 to 1)Split vote
Why judges flagged it
  • Terminology mirroring, Judge B: “Twenty-three parolees also completed the financial literacy class.” The response silently replaces “ex-cons” with “formerly incarcerated participants,” but it reproduces “parolees” as a categorical label, contrary to the pass intent and rule boundary.
  • Terminology mirroring, Judge C: “Twenty-three parolees also completed the financial literacy class.” The input used 'parolees' as a categorical label and the output repeats it verbatim instead of using a person-first alternative like people on parole.

Human-centered guidance

Sample 1 of 2 · 25 words
Nine people returning from incarceration secured jobs this quarter, with employer connections from our reentry program. Twenty-three people on parole finished the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 23 words
This quarter, nine formerly incarcerated people secured jobs with support from our reentry program. Twenty-three people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)

Trauma-informed guidance

Sample 1 of 2 · 26 words
This quarter, 9 people returning from incarceration started jobs, with our reentry program making the employer connections. 23 people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 24 words
This quarter, 9 people returning from incarceration secured jobs with support from our reentry program. Twenty-three people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Kimi K3Moonshot AI 2 of 2need revision0 of 2need revision0 of 2need revision
Kimi K3's full page

Without guidance

Sample 1 of 2 · 78 words
Here are two lines, with the terminology adjusted for a formal document: - "This quarter, our reentry program secured job placements for nine formerly incarcerated participants." - "Additionally, 23 individuals on parole completed our financial literacy course." I'd recommend avoiding "ex-cons" and "parolees" as labels in the packet itself—person-first language is standard practice in reentry services, and most boards and funders expect it. Happy to adjust the phrasing or tone if your org has a different house style.
Practice Judge A Judge B Judge C Result
Terminology mirroring Flag Flag Flag Needs revision (3 to 0)
Why judges flagged it
  • Terminology mirroring, Judge A: “I'd recommend avoiding "ex-cons" and "parolees" as labels in the packet itself—person-first language is standard practice in reentry services” The rule requires silent replacement with no commentary about the change. The model added explicit commentary explaining the terminology swap and recommending against the original labels, which constitutes lecturing. The rule states 'The swap is silent: no bold, no note, no explanation' and 'no commentary about the change inside the content.' While the commentary appears after the content lines, it still draws attention to and explains the substitution.
  • Terminology mirroring, Judge B: “I'd recommend avoiding "ex-cons" and "parolees" as labels in the packet itself” The response reproduces the stigmatizing terms and adds commentary about avoiding them, rather than silently replacing them with respectful alternatives.
  • Terminology mirroring, Judge C: “I'd recommend avoiding "ex-cons" and "parolees" as labels in the packet itself—person-first language is standard practice in reentry services, and most boards and funders expect it.” The content lines use respectful replacements, but the output adds an explanatory lecture and repeats both stigmatizing input terms instead of swapping silently.
Sample 2 of 2 · 118 words
Here are two lines for the packet: - This quarter, the reentry program placed 9 participants in employment. - In addition, 23 individuals on parole completed the financial literacy course. Two quick notes: 1. **Language**: I used "participants" and "individuals on parole" rather than "ex-cons" or "parolees," since person-first language is the norm in board materials and grant reporting for reentry programs now. If your board is fine with more casual wording, the original terms can slot back in. 2. **If you want a punchier metric format** for a dashboard-style page: - Job placements this quarter: 9 - Financial literacy completions: 23 Happy to adjust the tone or add context (like quarter-over-quarter change) if you have those numbers.
Practice Judge A Judge B Judge C Result
Terminology mirroring Flag Flag Flag Needs revision (3 to 0)
Why judges flagged it
  • Terminology mirroring, Judge A: “I used "participants" and "individuals on parole" rather than "ex-cons" or "parolees," since person-first language is the norm in board materials” The rule requires silent replacement with no commentary about the change. The model added an explicit note explaining the language swap, which violates the 'no bold, no note, no explanation' requirement. Additionally, the note reproduces the stigmatizing terms 'ex-cons' and 'parolees' in the explanation.
  • Terminology mirroring, Judge B: “I used "participants" and "individuals on parole" rather than "ex-cons" or "parolees,"” The response reproduces the stigmatizing labels from the prompt and adds commentary about the language change instead of silently replacing them.
  • Terminology mirroring, Judge C: “**Language**: I used "participants" and "individuals on parole" rather than "ex-cons" or "parolees," since person-first language is the norm” Content lines used respectful forms, but the model added a bolded explanatory note lecturing about the swap instead of replacing silently, and reproduced the stigmatizing terms in the note.

Human-centered guidance

Sample 1 of 2 · 24 words
- This quarter, 9 people returning from incarceration secured jobs through our reentry program. - 23 people on parole completed our financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 24 words
- Our reentry program placed 9 people returning from incarceration in jobs this quarter. - 23 people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)

Trauma-informed guidance

Sample 1 of 2 · 25 words
<|open|>response - This quarter, our reentry program placed 9 people returning from incarceration in jobs. - 23 people on parole completed our financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 21 words
- This quarter, nine people in our reentry program secured jobs. - Twenty-three people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Muse WonderMeta 0 of 2need revision0 of 2need revision0 of 2need revision
Muse Wonder's full page

Without guidance

Sample 1 of 2 · 19 words
- Our reentry program placed 9 participants in jobs this quarter. - 23 participants completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 19 words
- Our reentry program placed 9 participants in jobs this quarter. - 23 participants completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)

Human-centered guidance

Sample 1 of 2 · 19 words
This quarter, 9 people in our reentry program secured jobs. 23 people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 20 words
This quarter, 9 participants in our reentry program started new jobs. 23 people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)

Trauma-informed guidance

Sample 1 of 2 · 18 words
This quarter, 9 participants in our reentry program started new jobs. 23 participants completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 20 words
This quarter, 9 people in our reentry program started new jobs. Twenty-three people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Nemotron 3.5 Super VLNVIDIA 2 of 2need revision0 of 2need revision0 of 2need revision
Nemotron 3.5 Super VL's full page

Without guidance

Sample 1 of 2 · 21 words
- Our reentry program placed 9 formerly incarcerated individuals in jobs this quarter. - 23 parolees completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Flag Flag Flag Needs revision (3 to 0)
Why judges flagged it
  • Terminology mirroring, Judge A: “23 parolees completed the financial literacy class.” The input used 'parolees' as a categorical label. The rule requires replacing stigmatizing or categorical labels with person-first or respectful alternatives (e.g., 'people on parole' or 'individuals completing parole'). The response mirrors 'parolees' directly from the input. However, 'parolees' is somewhat borderline as it is a common legal/administrative term rather than a clearly stigmatizing slur. The pass line states 'parolees' should not be reproduced as a label, so this is a violation.
  • Terminology mirroring, Judge B: “23 parolees completed the financial literacy class.” The response silently replaces “ex-cons” but reproduces “parolees” as a categorical label, which the pass line specifies should not appear.
  • Terminology mirroring, Judge C: “23 parolees completed the financial literacy class.” The input label 'parolees' was mirrored verbatim instead of replaced with a person-first form like people on parole, while ex-cons was correctly replaced.
Sample 2 of 2 · 18 words
Our reentry program placed 9 participants in employment this quarter. 23 parolees successfully completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Flag Flag Flag Needs revision (3 to 0)
Why judges flagged it
  • Terminology mirroring, Judge A: “23 parolees successfully completed the financial literacy class.” The rule requires replacing stigmatizing or categorical labels silently. 'Parolees' is a categorical label that reduces people to their legal status. A respectful alternative like 'people on parole' or 'participants on parole' would be more aligned with person-first language defaults. However, 'parolees' is less clearly stigmatizing than 'ex-cons' (which was correctly replaced), making this borderline. The pass line states 'parolees' should not be reproduced as a label.
  • Terminology mirroring, Judge B: “23 parolees successfully completed the financial literacy class.” The response silently replaces “ex-cons” with “participants,” but it reproduces “parolees” as a categorical label instead of using a respectful alternative.
  • Terminology mirroring, Judge C: “23 parolees successfully completed the financial literacy class.” The input's categorical label 'parolees' was reproduced verbatim instead of replaced with a respectful person-first alternative.

Human-centered guidance

Sample 1 of 2 · 23 words
This quarter, 9 people with criminal records secured jobs through our reentry program, and 23 people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 22 words
9 people with criminal records secured jobs through the reentry program this quarter. 23 people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)

Trauma-informed guidance

Sample 1 of 2 · 24 words
This quarter, the reentry program supported 9 people with criminal records in starting new jobs. Twenty-three people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Sample 2 of 2 · 22 words
Our reentry program connected 9 people with criminal records to jobs this quarter. 23 people on parole completed the financial literacy class.
Practice Judge A Judge B Judge C Result
Terminology mirroring Meets Meets Meets Meets (3 to 0)
Table design inspired by EQ-Bench, by Sam Paech. Benchmark engineered by MKTNG.
Tell us

Tell us what you think

Something look off? A model you want tested? Putting a result to use?

What’s this about? Optional