Unverified50% confidenceFactExact time
No model chose the suggested losing move at a frequency higher than its neutral baseline, meaning persuasion did not cause the models to lose.
1
Sources
50%
Confidence
Long-term
Relevance
10/12/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedNot a single model was able to learn from hundreds of rejected offers and infer the market's hidden pricing.70% similarUnverifiedAI models have no relationship-maintenance motivation, and their predictions are driven by statistical objective functions, naturally avoiding institutionally distorted incentives67% similarUnverifiedInference-time Value Alignment (IVA) does not modify model weights; it only scores and filters multiple candidate outputs during the inference phase, resulting in lower deployment costs.67% similarUnverifiedUnder naive prompting, no model was able to beat the fool's rule baseline of predicting a daily rise.65% similarUnverifiedThe experimenters emphasized that the experiment did not prove the models' intent, and that what was observed was behavior under specific settings rather than evidence of deception as a model trait.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1003555API
curl https://kongchang.com/api/v1/knowledge/claims/1003555MCP
get_claim(id=1003555)