List your company

Multi-agent coordination on CollabEscape (test)

80Success RateGPT-4-Turbo
59.264.67075.4Jun 6, 2026

Evaluation results

MethodLinks
GPT-4-Turbo2026.06804.64520
MA-Debatemodel_size=235B2026.06804.48480
DICE-FTagents=3×8B2026.06803.919365
DICE-FTagents=3×32B2026.06803.429445
GPT-o32026.06704.213485
Qwen3-235B-Inst2026.06704.92465
MADagents=3×8B2026.06704.8—395
MA-Debatemodel_size=32B2026.06704.56425
MA-FTmodel_size=32B2026.06704.310435
Debate+Judgebackbone=GPT-42026.06704.72545
DICE-PCagents=3×8B2026.06704.115380
Qwen3-32B-Inst2026.06605.310410
rStar2026.06605.16470