Strategic Intelligence in Large Language Models: Evidence from evolutionary Game Theory · HackerPapers discussions

Strategic Intelligence in Large Language Models: Evidence from evolutionary Game Theory — mentioned by code_biologist

Wed, 19 Aug 2026 19:02:56 GMT

The personality of the underlying model persists even if you're able to alter surface tone enough for your needs. An observation from a year ago: people on AI text RPG subreddits saying they had a lot of difficulty with Gemini role playing ambiguous characters and that almost always the characters would betray them, or misread the human role player's motives as negative. Someone pointed out this paper [1] that showed significant differences between the different LLMs strategic behavior playing iterated prisoner's dilemma where Gemini had exactly that behavior, and speculatively that…

Paper and discussion history. A mention is not an endorsement.

Strategic Intelligence in Large Language Models: Evidence from evolutionary Game Theory — HN submission

Mon, 14 Jul 2025 17:41:39 GMT

Strategic Intelligence in LLMs: Evidence from Evolutionary Game Theory

2 points; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.

Strategic Intelligence in Large Language Models: Evidence from evolutionary Game Theory — HN submission

Fri, 11 Jul 2025 14:22:12 GMT

Strategic Intelligence in Large Language Models

1 point; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.

Strategic Intelligence in Large Language Models: Evidence from evolutionary Game Theory — HN submission

Tue, 08 Jul 2025 07:03:18 GMT

Strategic Intelligence in Large Language Models: Evidence from Evolutionary GT

2 points; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.