Evaluation trends,
distilled week by week.
Each issue captures the week’s most noteworthy benchmarks and datasets, along with the methodological shifts they reveal.
From question answering to systems in motion
FDE-Gym · DataWorld-Bench · LADDER · MindEval2 · Realtime-Bench · EvoWeb · WorldMemBench · Grid-Bench · J-tau
PICKS↗W41
Executable environments, process audits and real usability
BrickBench · SpaceCast-Bench · SWE-CC · DUDA-Bench · TaoD2C-Bench · TopoGraphRAG-Bench · SAVU-Bench · Learn2Play Bench · WorldFact-Bench
PICKS↗W38
From static questions to executable, replayable evaluation
ScienceIDE · ProgramDistill · VA-Bench · BVB · ImpossibleRubrics · RiskChainBench · HearInContext · VoiceTrace-Bench · RoleBreak
PICKS↗W37
Agent benchmarks become evaluation systems
Harbor Adapters + Harbor-Index · Mr.LHDR · MOLE · SchemeArena · IdeaAMBIG · HLS-Eval Agentic Flow · TransClean · CausalArena · SynthGait-19K
PICKS↗W36
Agent evaluation becomes infrastructure
Terminal-Universe · HarnessDev · Last Translation Benchmark · EarlyEval · TIC-Bench · Principia · DroneCATS · RoboTok · SonicCaps
PICKS↗W35
State, process and environment execution
MobilePA-Bench · EarthVerse · MCR-Bench · SWE Refactor Bench · Modality Maturity Index (MMI) · LAION-BVD · CorporateBench · TurnBench · MechBench-36
PICKS↗W34
Real workflows, persistent state and executable verification
ASI-Bench · StartupBench · SWE-bench Science · StateMemBench · CHIVE · HarnessRisk · HiPHI · ReRef-3D · RuleMaze
PICKS↗W33
Discovery, tool use and long-horizon agents
DiG-bench · VAKRA · FrontierFinance · LigBench · LongEarth-Bench · MBA-Bench · QuoteBench · SWE-Bench ProMax · VibeLifeBench
PICKS↗
The next issue is being curated
Scheduled for Sunday