Skip to content
Apart Research
Apart Research
Contact us
Sprint projectMay 25, 2026Montreal

JARE: A Differential-Testing Workbench for Auditing AI-Generated Specifications

Allenna Tang, Jerry Chen, Elizabeth Kourbatski · Team JARE

Submitted to The Secure Program Synthesis Hackathon. Sprint projects are early-stage work by participants, not Apart Research publications.

Read the report

Report: JARE: A Differential-Testing Workbench for Auditing AI-Generated Specifications

Open in a new tab (opens in new tab)
Recording (opens in new tab)Code (opens in new tab)
Share

JARE is an auditing tool for AI-generated software specifications. When an AI writes both the rules a program should follow and the tests that check those rules, the tests can pass even when the rules are wrong. JARE finds concrete examples where an AI-generated specification disagrees with what the developer actually wanted, so a human can review and fix it.

Reviews

Judging this Sprint?

Review this project

Your public critique appears on this page without your name. Your private critique is not published; only the Apart team reads it. If you agree below, we share your review with grantmaking.ai (opens in new tab) and the Transformative AI Fund so strong projects can be funded.

Not shown on this page.

Shown on this page, without your name.

Only the Apart team reads this, and funders if you agree below.

Share my name publicly on grantmaking.ai *
Share my private critique with funders *

How much would this matter for AI safety if it worked? How innovative is it? For scores of 4-5: is this actually new to the field, or replicating recent work?

Scoring guide
  1. 1Negligible. No clear problem addressed, or no meaningful novelty.
  2. 2Limited. Addresses a real problem but with a generic or well-trodden approach. Incremental at best.
  3. 3Moderate. Clear problem with a reasonable approach; some novelty in framing or method beyond routine application of existing tools.
  4. 4Significant. Important problem with an original approach, or identifies a neglected problem area. A valuable contribution others could build on.
  5. 5Exceptional. Tackles a critical AI safety problem with a genuinely novel approach, or opens a new research direction. Clear theory of change. You'd be excited to share this with researchers in the area.

How sound are methodology, implementation, and findings?

Scoring guide
  1. 1Seriously flawed. Methodology broken, results uninterpretable, or implementation doesn't work.
  2. 2Weak. Approach has significant gaps: missing validation, flawed experimental design, or incomplete implementation.
  3. 3Competent. Technically solid given the short duration. Methodology makes sense, results are interpretable, limitations acknowledged, work builds toward clear conclusions.
  4. 4Strong. Thorough methodology with convincing validation. Results clearly support conclusions. Immediately useful for future work.
  5. 5Exceptional. Ambitious scope executed rigorously. Surprising findings, novel methods, or unusually robust validation.

How clearly are work, findings, and impact potential communicated?

Scoring guide
  1. 1Incomprehensible. Cannot determine what the project is actually claiming or doing.
  2. 2Hard to follow. Key information buried, missing, or diluted by excessive length. Significant effort to extract main points.
  3. 3Clear enough. Can understand the problem, approach, and results without undue effort. Core content clearly present: problem, method, findings, limitations.
  4. 4Well presented. Easy to follow, well-structured, appropriate level of detail. Target audience would get it quickly.
  5. 5Exceptionally clear. A pleasure to read. Complex ideas made accessible. Could serve as a model for how to present this type of work.

  1. I'm afraid I can't spot a way that the proposed functionality is at all useful in program synthesis, vs. in carrying out evaluation of synthesis tools. The approach assumes collateral that would be very useful in the hands of an AI generating specs, yet the collateral is held out and only applied after generation.

    One proposed usage mode is comparing a generated spec against a known-good spec. But if you have a known-good spec, why are you using AI to generate a spec from scratch, given just a natural-language description? Shouldn't the AI at least be shown the known-good spec, too?

    Another proposed usage mode is evaluating generated specs against labeled test cases. Again, test cases are extremely useful to agentic AI-coding tools, so why aren't you sharing them?

    Regardless of which scenario we focus on, the proposed functionality seems to follow standard practice in systematic testing, with no new conceptual contribution.

    Read full reviewShow less

  2. Promising experiments in a useful direction. One improvement could have been to compare different models in a multi-model setting: because eg ChatGPT and Claude are trained on different data, they can to some degree compensate for each others weaknesses, which may help significantly improve results in the no-gold deployment setting.

Cite this project

@misc{tang2026jare,
  title = {{JARE: A Differential-Testing Workbench for Auditing AI-Generated Specifications}},
  author = {Allenna Tang and Jerry Chen and Elizabeth Kourbatski},
  year = {2026},
  month = may,
  note = {Submitted to The Secure Program Synthesis Hackathon, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/jare-a-differentialtesting-workbench-for-auditing-aigenerated-specifications-o86j}},
  url = {https://apartresearch.com/sprints/projects/jare-a-differentialtesting-workbench-for-auditing-aigenerated-specifications-o86j}
}

More from The Secure Program Synthesis Hackathon

All 61 entriesBrowse all projects

Build something like this at the next Sprint

AI Collusion Research Sprint · Oct 23 - 25, 2026