Skip to content
Trending storyDeveloping

Stanford study finds more shared information helps some LLM trading agents and hurts others

1 article1 sourcesince Nov 7Last article 43m ago ·

Overview

AISummary of 1 article

A Stanford study tested 10 frontier LLMs in trading, bargaining and double auctions, changing only how precisely each agent's private values were shared and scoring gains against known utilities.

The effect differed by model. Exact values cut GPT-6 Sol's share of maximum gains in trade from 98% to 91%, while Claude Sonnet 5's rose from 72% to 86%, according to a report of the study by Rohan Paul on X. In auctions, full disclosure inflated price demands and lowered Grok 4.7's share from 97% to 62%.

Written by AI from the articles below · updated Nov 7, 5:17 PM ET

Check the sources:

Latest developmentNov 7, 5:05 PM ET
Stanford study finds more information helps some LLM agents trade and hurts others

Article timeline

The articles in this story. Times are ET.

Nov 7
  1. Rohan PaulX
    Stanford study finds more information helps some LLM agents trade and hurts others

    AIA Stanford study tested 10 frontier LLMs in trading, bargaining and double auctions, varying how precisely private values were shared. Exact values cut GPT-6 Sol's share of maximum gains in trade from 98% to 91% but raised Claude Sonnet 5's from 72% to 86%. In auctions, full disclosure inflated price demands and dropped Grok 4.7 from 97% to 62%.

    Image from @rohanpaul_ai's post

Heat Trend

Not enough continuous observations to show a trend yet.