New method Nous extracts but fails to inject human cognition into LLM agents

Researchers have developed a method called Nous to extract and inject cognitive profiles from human traders into LLM agents operating in prediction markets. The study found that while it's possible to extract stable behavioral parameters from trading activity, injecting these profiles into agents via prompts did not measurably improve their performance or reduce forecast correlation. The limitations appear to stem from the prompt-level injection method, suggesting that deeper integration techniques like fine-tuning or activation steering may be necessary to effectively transfer human cognitive diversity to AI agents. AI

IMPACT Suggests prompt-level injection is insufficient for transferring human cognitive diversity to AI agents, motivating research into deeper integration methods.

RANK_REASON Academic paper detailing a novel method and its experimental results. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• LLM agents
• Nous
• Polymarket
• prediction markets

• paper
• model release
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →