keystone-bench

A clinical decision-evidence benchmark: pull the one fact an answer rests on, and see whether the assistant's action moves with the evidence. Built on HealthBench.

xinxuxin/keystone-bench

View on GitHub ↗

A clinical decision-evidence benchmark: pull the one fact an answer rests on, and see whether the assistant's action moves with the evidence. Built on HealthBench.

keystone-bench grew +28.6% in the last 30 days, ranks 128th of 872 in Evaluation & Guardrails, last commit 33 days ago.

Star trajectory

68 days tracked
2026-09-141.1k stars today2026-11-22
Momentum

+28.6% in 30d, above the Evaluation & Guardrails median of +1.4%.

Maintenance

Last commit 33 days ago, staler than the median of 15d.

How it compares in Evaluation & Guardrails

keystone-bench
1.1k
stars
+28.6% 30d
commit 33 days ago
Evaluation & Guardrails median
235
stars
+1.4% 30d
commit ~15d ago
Leader · impeccable
114.7k
stars
+0.0% 30d
commit today

Ranked 128th of 872 by stars.

Ranked by shared tags, then keywords, then stars
  1. Dingo: A Comprehensive AI Data, Model and Application Quality Evaluation Tool

    +10 (+1.3%) in 30d · Committed 2d ago · Python · Apache-2.0

    Shares Evaluation & Guardrails, Dataset Engineering

  2. Open source release of the evaluation benchmark suite described in "Realistic Evaluation of Deep…

    +0 (+0.0%) in 30d · Committed 2,687d ago · Python · Apache-2.0

    Shares Dataset Engineering, Evaluation & Guardrails

  3. Benchmark for automated failure attributions in agentic systems (🏆 ICML 2025 Spotlight)

    -2 (-0.5%) in 30d · Committed 22d ago · Python · MIT

    Shares Evaluation & Guardrails, Dataset Engineering

  4. Code for the paper "Rethinking Benchmark and Contamination for Language Models with Rephrased Samples"

    -1 (-0.3%) in 30d · Committed 1,067d ago · Python · Apache-2.0

    Shares Dataset Engineering, Evaluation & Guardrails

  5. Official code for Agent-Reflex: Teaching Language Agents to Act Reflectively

    +20 (+8.4%) in 30d · Committed 33d ago · Python · MIT

    Shares Dataset Engineering, Evaluation & Guardrails

  6. A unified, downstream-grounded benchmark for LLM-driven training data construction, selection, and quality…

    +1 (+0.4%) in 30d · Committed 19d ago · Python

    Shares Dataset Engineering, Evaluation & Guardrails

  7. PatchEval: A New Benchmark for Evaluating LLMs on Patching Real-World Vulnerabilities

    +4 (+1.7%) in 30d · Committed 34d ago · Python · Apache-2.0

    Shares Dataset Engineering, Evaluation & Guardrails

  8. Autonomous Qwen3-VL training-code research on the official DocVQA benchmark. main: NVIDIA multi-GPU, mlx…

    +0 (+0.0%) in 30d · Committed 22d ago · Python · MIT

    Shares Evaluation & Guardrails, Dataset Engineering

  9. [EMNLP 2026] A multilingual benchmark for repository-level unit test generation, evaluating LLMs across 5…

    +1 (+0.6%) in 30d · Committed 86d ago · Python · Other license

    Shares Dataset Engineering, Evaluation & Guardrails

  10. A Comprehensive Benchmark for Routing LLMs to Explore Model-level Scaling Up in Large Language Models

    +0 (+0.0%) in 30d · Committed 172d ago · Python · MIT

    Shares Evaluation & Guardrails, Dataset Engineering

Language
Python
License
Other
Contributors
1
Forks
3
Watchers
1
Open issues
0
Releases
2
Created
2026-09-07
Stars this week
-1

You've seen where keystone-bench stands today.

Get the moves before everyone else: the 5 Evaluation & Guardrails repos climbing fastest, every Tuesday.

Free weekly digest. Double opt-in, unsubscribe anytime.

Maintain keystone-bench?

Claim this page to add your own call-to-action.

Claim this page