Does Learning to Predict the World Help Agents Act? Auditing World-Model Post-Training

arXiv:2609.33335 · cs.CL, cs.AI · Submitted 2026-09-27 · Read on arXiv

cs.CL, cs.AI

Submitted: 2026-09-27

Updated: 2026-09-27

Code: https://github.com/KosmoCHE/WM-PostTraining-Audit

Terminology

• GRPO — 30× in this paper · explained in 6 episodes
• Accuracy — 27× in this paper · explained in 1 episode
• World Model — 17× in this paper · explained in 7 episodes
• Distillation — 10× in this paper · explained in 2 episodes
• Self-distillation — 6× in this paper · explained in 6 episodes
• Benchmark — 5× in this paper · explained in 2 episodes
• Trace — 5× in this paper · explained in 2 episodes
• Benchmarking — 4× in this paper · explained in 3 episodes
• Core — 4× in this paper · explained in 2 episodes
• Regularization — 3× in this paper · explained in 3 episodes
• Reinforcement Learning — 3× in this paper · explained in 4 episodes
• Adam — 2× in this paper · explained in 2 episodes
• Cosine Similarity — 2× in this paper · explained in 3 episodes
• Hallucination — 2× in this paper · explained in 3 episodes
• KL Regularization — 2× in this paper · explained in 2 episodes
• AI Agents — 1× in this paper · explained in 2 episodes
• Chain-of-Thought (CoT) — 1× in this paper · explained in 4 episodes
• LLM judge — 1× in this paper · explained in 2 episodes
• On-policy Self-Distillation (OPSD) — 1× in this paper · explained in 2 episodes
• Reasoning Traces — 1× in this paper · explained in 3 episodes
• Steering — 1× in this paper · explained in 2 episodes

Sources

• PatchWorld: Gradient-Free Optimization of Executable World Models for Agent Environments
• Qwen2.5-VL Technical Report
• Beyond Next-Observation Prediction: Agent-Authored World Modeling for Sequential Decision Making
• Evaluating Large Language Models Trained on Code
• The Llama 3 Herd of Models
• Better World Models Can Lead to Better Post-Training Performance
• Dream to Control: Learning Behaviors by Latent Imagination
• Privileged, but Biased: How PI-Conditioned Teachers Break Self-Distillation
• Privileged Solutions or Context-Induced Teacher Behavior? Dissecting On-Policy Self-Distillation
• Form, Not Content? A Preregistered, Placebo-Controlled Evaluation of Learned Error-Conditioned Self-Repair Through Prompts and Weights in Frozen Small Code Models
• Why Does Self-Distillation (Sometimes) Degrade the Reasoning Capability of LLMs?
• TAPO: Transition-Aware Policy Optimization for LLM Agents
• CoMAP: Co-Evolving World Models and Agent Policies for LLM Agents
• Policy and World Modeling Co-Training for Language Agents
• Describe-Then-Act: Proactive Agent Steering via Distilled Language-Action World Models
• Qwen2.5 Technical Report
• DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
• ECHO: Terminal Agents Learn World Models for Free
• Thinking by Doing: Building Efficient World Model Reasoning in LLMs via Multi-turn Interaction
• SWE-World: Building Software Engineering Agents in Docker-Free Environments

Related papers

• Exploring Solution Divergence and Its Effect on Large Language Model Problem Solving
• Ishigaki-IDS-Bench: A Benchmark for Generating Information Delivery Specification from BIM Information Requirements
• Subliminal Steering: Stronger Encoding of Hidden Signals
• MedStruct-S: A Benchmark for Key Discovery, Key-Conditioned QA and Semi-Structured Extraction from OCR Clinical Reports
• The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures
• Untangling the Mechanisms of Misleading Context in Medical Question Answering