Multi-agent RL post-training fights LLM mode collapse and boosts response diversity
natashajaques · x · 2026-09-16
LLMs give eerily similar answers across model families — ask six times for a "well-received book" and it's always To Kill a Mockingbird.
Natasha Jaques et al. propose multi-agent RL post-training as a fix:
- The same model adopts different roles that learn to diversify responses, but only above a quality threshold
- Outperforms several diversity-enhancing baselines on 4 benchmarks spanning scientific ideation and creative writing
- Framed by an evolution quote: nature makes every individual unlike every other, while LLM mode collapse forces homogeneity
A method aimed directly at the root of AI-assisted creativity's sameness problem.
Related event: Multi-Agent RL Tackles LLM Mode Collapse(2 posts)→
More from Research
- Robotics researcher pushes back on 'omni embodiment' hype: it's the hands, not the abstraction — chris_j_paxton · 2026-09-16
- Mathematician Tony Feng: current AI is not robustly superhuman yet — littmath · 2026-09-16
- Research team builds open science-task taxonomy from job postings — JMateosGarcia · 2026-09-16
- Anthropic researcher, aided by Claude, cracks Classic McEliece challenge instance — matthew_d_green · 2026-09-16
- StarVLA's VLAct trains VLAs on 16 GPUs by reshaping action representations, not data scaling — jiqizhixin · 2026-09-16
- RL Gains Concentrate on Easy Questions; 'Never Give Up' Resampling Tackles the Hard Ones — teortaxesTex · 2026-09-16