LIVE PULSE
4.8 Anthropic CEO Amodei calls for slower AI development and shared safety rules — 11 src2.6 Agility Robotics unveils Digit 5 humanoid for warehouses and factories — 2 src2.4 Apple ships rebuilt Siri with Google Gemini, but not in the EU — 2 src2.1 Siri AI in macOS 27 Golden Gate: FAQ, Germany availability, privacy questions — 2 src1.7 Sam Altman says OpenAI will not go public in 2026, citing AI safety concerns — 5 src1.4 OpenAI contractors review real ChatGPT conversations to rate responses, report says — 2 src1.3 Anthropic data retention policy prompts firms to limit Claude use for sensitive work — 1 src1.3 Paper Examines How First Query Shapes Agentic Deep Search — 1 src1.3 Conformance-Driven Iterative Refinement for Natural-Language to SysMLv2 Translation — 1 src1.3 Neural Operators for Nonlinear Functionals on RKHS — 1 src4.8 Anthropic CEO Amodei calls for slower AI development and shared safety rules — 11 src2.6 Agility Robotics unveils Digit 5 humanoid for warehouses and factories — 2 src2.4 Apple ships rebuilt Siri with Google Gemini, but not in the EU — 2 src2.1 Siri AI in macOS 27 Golden Gate: FAQ, Germany availability, privacy questions — 2 src1.7 Sam Altman says OpenAI will not go public in 2026, citing AI safety concerns — 5 src1.4 OpenAI contractors review real ChatGPT conversations to rate responses, report says — 2 src1.3 Anthropic data retention policy prompts firms to limit Claude use for sensitive work — 1 src1.3 Paper Examines How First Query Shapes Agentic Deep Search — 1 src1.3 Conformance-Driven Iterative Refinement for Natural-Language to SysMLv2 Translation — 1 src1.3 Neural Operators for Nonlinear Functionals on RKHS — 1 src
HEATPULSEAI MAGAZINES
FLIP · FOLLOW · SAVE
Log inSign up

ContractEval

model1 events
papersSEP 10 04:00 UTC

ContractEval benchmark checks whether LLM agents follow procedural instructions correctly

Researchers introduced ContractEval, a benchmark that evaluates whether LLM agents actually perform the required steps when carrying out a procedure rather than just producing a plausible answer. It uses query-conditioned execution matching to surface hidden failures where a required check, branch, dependency, or invariant was omitted despite the final output appearing correct. The work targets gaps in output-only evaluation for agentic systems.

agents/benchmark/llm/evaluation