List your company

Skill-selection on ToolBench-60 soft-0.3 preference regime (held-out pool)

255.9RegretBandit-as-Override
246.468310.134373.8437.466Jun 4, 2026

Evaluation results

MethodLinks
Bandit-as-OverrideBackbone=DeepSeek-V4-Flash, Agent Category=LLM with Statistical Prior2026.06255.947.20.523
Bandit-as-OverrideBackbone=GPT-5.22026.06256.847.30.535
Pure-BanditBackbone=GPT-5.22026.06279.240.30.554
Pure-BanditBackbone=DeepSeek-V4-Flash, Agent Category=Statistical2026.06280.739.80.549
Freq-as-OverrideBackbone=GPT-5.22026.06286.5430.286
Freq-as-OverrideBackbone=DeepSeek-V4-Flash, Agent Category=LLM with Statistical Prior2026.0628841.60.273
Bandit-as-ContextBackbone=DeepSeek-V4-Flash, Agent Category=LLM with Statistical Prior2026.06309.6370.465
Profile-MemoryBackbone=GPT-5.22026.0631638.60.193
Bandit-as-ContextBackbone=GPT-5.22026.06327.1350.473
Freq-GreedyBackbone=GPT-5.22026.06327.2340.404
Profile-MemoryBackbone=DeepSeek-V4-Flash, Agent Category=LLM with Memory2026.06327.434.80.251
Freq-GreedyBackbone=DeepSeek-V4-Flash, Agent Category=Statistical2026.0632833.70.402
InCotext-MemoryBackbone=DeepSeek-V4-Flash, Agent Category=LLM with Memory2026.06372.525.50.294
InCotext-MemoryBackbone=GPT-5.22026.06373.525.20.304
ZeroShot-LLMBackbone=GPT-5.22026.06375.524.80
ZeroShot-LLMBackbone=DeepSeek-V4-Flash, Agent Category=No Learning2026.06376.724.60
RandomBackbone=GPT-5.22026.06491.51.60
RandomBackbone=DeepSeek-V4-Flash, Agent Category=No Learning2026.06491.71.60