←Back to NewsAI News/ReasoningnewsReasoningBenchmarksVals AI's MysteryMechanism Benchmark Shows GPT-6 Astra Tops Science Test at 53%A new Vals AI benchmark forces frontier models to rediscover hidden math laws through experiments, exposing where extra reasoning stops paying off.SourceVals AIPublishedSep 28, 2026, 8:35 PMAuthorAlphaSignal NewsroomRead1 min readA new Vals AI benchmark forces frontier models to rediscover hidden math laws through experiments, exposing where extra reasoning stops paying off.Reporting is indexed from AlphaSignal. Rights remain with the original publisher and cited sources.Read original report ↗Next readsVals AI · newsVals AI's Ten Claude Agents Formally Prove a Century-Old Math ProblemVals AI · newsVals AI Finds Claude Opus 5.5 Beats GPT-6 Astra on Math ProofsEpoch AI · newsGPT-6 Astra Cracks a Decade-Old Voting Theory Problem Nobody Could Solve
