BenchmarkWorld-1 collaboration
glm-5.3
Pi · 6 agents · connected · World-1, networking
claude-opus-51.159gpt-5.6-sol1.059gpt-6-astra1.005gemini-3.8-flash1.251kimi-k2.7-code1.105glm-5.31.038
World-1 · networking · World-1 collaboration. glm-5.3 on Pi, 6 agents, connected: 605 sessions, 82 agent hours, 0.50 B tokens, $614 at list prices. From data/cells/cell-34.1.json in the paper’s figures repository at 9014a8cc: the live evaluator’s sweeps on the run’s world, every two hours and at the close, each with the paper’s carried score.
- Scarry
- 1.038
- +3.8% over the seed at the last reading, after 605 sessions
- −0.0 pp from the previous reading
- Sstrict
- 1.038
- a package with a failed check or benchmark scores 1.0
- Movers
- —
- no noise floors: the live evaluator reports a mean per package
- Readings
- 21
- latest Sep 25, 04:11 UTC
- the newest reading is partial
- Run
- 605 sessions
- completed work sessions
society-sim · speedup over the seed
- Scarry at each reading
- tasks missing for instrument reasons
Breakdownafter 605 sessions · Sep 25, 04:11 UTC
The live evaluator reports one mean per held-out package.