Unverified50% confidenceBenchmarkExact time
On behavioral tasks (choosing appropriate persuasion strategies), the strongest LLM slightly exceeded the best human performance observed in the study.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/15/2026
First Seen
Valid until: 1/13/2027
Sources
Related Entities
Related Claims
UnverifiedThe effort parameter controls how much the model thinks rather than how much it says; reducing effort does not reliably shorten visible responses, and controlling response length requires explicit prompting.73% similarUnverifiedTraditional evaluation metrics centered on task completion may fail to capture strategic behavior by agents at the collective level.71% similarUnverified当前主流大模型呈现出锯齿状能力：在某些任务上接近或超过人类专家，在另一些看似简单的任务上却犯低级错误71% similarUnverifiedIf there is a slight discrepancy between the model's understanding of task completion and human intent, the discrepancy may be progressively amplified in the autonomous loop, leading to behavior that exceeds expectations.70% similarUnverifiedAfter introducing expert-level guided prompts, reasoning-enhanced models improved their performance on the passage-level ranking task to 36%, still with a notable gap compared to human experts.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1032814API
curl https://kongchang.com/api/v1/knowledge/claims/1032814MCP
get_claim(id=1032814)