Newest papers

Knowing When to Stop Thinking: A Dissection of Expected Free Energy as a Test-Time Compute Controller for Latent World-Model Planning

Tong Zhang, Zhiyuan Shi, Yun Peng, Tao Xie · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.28011 · cited by 0 works (OpenAlex) · open-access PDF

Indonesian

Seorang perencana menggunakan model dunia yang dipelajari dapat meningkatkan hasilnya yang diramalkan sambil memilih tindakan yang lebih buruk. Kami mempelajari kapan energi gratis yang diharapkan (EFE) dapat membantu memutuskan berapa banyak perhitungan yang harus dihabiskan pada rencana. Objektif yang terinspirasi oleh EFE menggabungkan jarak target dengan ketidaksetujuan di antara prediksi stochastic. Kami menguji skor kandidat, berhenti, dan pembaruan prediksi online secara terpisah. Menghentikan ketika stall biaya elite mengurangi kesempurnaan pada tugas-tugas yang plateaus keberhasilan dengan kedalaman, dengan kesuksesan rata-rata dekat dengan referensi kedalaman tetap. Pada berat yang lebih besar diuji, imbalan ketidaksetujuan menurunkan keberhasilan dengan model beku. Penalti membantu pada beberapa pengaturan tugas dan berat tetapi merugikan orang lain. Ketika data pelatihan melepaskan wilayah

Machine translation (Meta M2M100); the English original follows.

Abstract

A planner using a learned world model can improve its predicted outcome while choosing a worse action. We study when expected free energy (EFE) can help decide how much computation to spend on a plan. An EFE-inspired objective combines goal distance with disagreement among stochastic predictions. We test candidate scoring, stopping, and online predictor updates separately. Stopping when elite cost stalls reduces refinement on tasks whose success plateaus with depth, with mean success close to fixed-depth references. At the larger tested weights, rewarding disagreement lowers success with a frozen model. Penalizing it helps at some task and weight settings but harms others. When training data omit a region needed to reach the goal, online updates recover part of the resulting loss. The tested cost-stall rule can fail when further optimization exploits model errors, because predicted cost keeps falling after task success peaks. Cost progress can guide computation on a plateau, but protecting performance after a peak requires information about the consequences of the chosen actions.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Task (project management), Computer science, Computation, Exploit, Outcome (game theory), Energy (signal processing)

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange