BELIEFWEAVER improves long-horizon agent recovery from earlier failures
A distilled 9B student gained 9.8 points over the native baseline and cut estimated teacher API cost by 54.8%.
2 stories, newest first · one card per story
A distilled 9B student gained 9.8 points over the native baseline and cut estimated teacher API cost by 54.8%.
Probing six open-source LLMs on HMM-generated data shows belief states are linearly decodable (R² 0.83–0.99), suggesting in-context learning approximates Bayesian prediction.
Today 1.6K readers · 2.5K views · 778 reads · all time 189K views