Self-Organizing Agent Teams Learn to Reason Together
Self-organizing agent teams learn collaboration strategies and beat their strongest member on math and physics benchmarks.
Self-Organizing Agent Teams are fixed groups of AI agents that learn reusable strategies for roles, conversational phases, participation, and information flow from prior collaborations. Strategies learned from only 15 mathematics and 25 graduate-level problems transfer unchanged to unseen benchmarks. Across five math and physics benchmarks the teams average 66.7% accuracy, versus 48.8% for the strongest member and 59.0% for a perfect router, and they exceed that router by 13.4 points on AIME 2026. Demonstrability of correct reasoning strongly predicts collaboration gains, with Spearman ρ of 0.90.
- Teams trained on 40 problems transfer strategies to unseen benchmarks.
- Average accuracy is 66.7% versus 48.8% for the strongest member.
- On AIME 2026 they exceed a perfect router by 13.4 points.
- Demonstrability tracks gains over the strongest member with Spearman ρ of 0.90.
Full article265 words · extracted from huggingface.co · click to collapse
Collective intelligence depends not only on what team members know, but also on how they organize their work. When the structure of a solution is unknown, useful roles and divisions of labor cannot be specified in advance; teams must learn from experience how to organize reasoning as it unfolds. Human teams routinely adapt this way, while existing AI agent teams rely on fixed protocols, explicit task decomposition, or routing. We introduce Self-Organizing Agent Teams (SAT), fixed teams of AI agents that learn reusable strategies from prior collaborations to organize roles, conversational phases, participation, and information flow. These strategies enable what we call collaborative computation: agents exchange, challenge, repair, and synthesize partial reasoning into solutions no member produced independently. In two independent settings, we learn teamwork strategies that transfer unchanged to unseen benchmarks, using only 15 mathematics and 25 graduate-level knowledge problems. Across five mathematics and physics benchmarks, self-organizing teams average 66.7% accuracy, versus 48.8% for their strongest member, 58.7% for compute-matched inference by that agent, and 59.0% for a perfect router over members' independent answers; on AIME 2026, they exceed this router by 13.4 points. Because gains vary across benchmarks, we ask when self-organizing collaboration helps. Across eight benchmarks, demonstrability (the organizational-psychology construct of whether a team can distinguish correct from incorrect reasoning) strongly tracks improvement over the strongest member (Spearman ρ=0.90, p=0.005): teams benefit most when correct reasoning can be recognized once it appears. More broadly, these results suggest that organization itself can become an agent capability: agent teams can learn how to reason together and produce solutions their members could not reach independently.
Text extracted automatically; images, tables and formatting may be missing. Original: https://huggingface.co/papers/2609.22682