From Imitation to Reward Discovery: On-Policy Warmup for Agentic RL
cs.LG, cs.AI
Submitted: 2026-09-30
Updated: 2026-09-30
Sources
- $\tau^2$-Bench: Evaluating Conversational Agents in a Dual-Control Environment
- The Coverage Principle: How Pre-Training Enables Post-Training
- Evaluating Large Language Models Trained on Code
- DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
- RL Starts before RL: On Policy Distillation for Better Reinforcement Learning
- Distilling the Knowledge in a Neural Network
- Sequential Beats Joint: On the Interplay between On-Policy Distillation and RLVR
- On-Policy Distillation with Curriculum Turn-level Guidance for Multi-turn Agents
- ReFT: Reasoning with Reinforced Fine-Tuning
- TailSFT: Filtered Fine-Tuning Improves Post-Training Performance
- DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
- ALFWorld: Aligning Text and Embodied Environments for Interactive Learning
- Beyond SFT-to-RL: Pre-alignment via Black-Box On-Policy Distillation for Multimodal RL
- OpenClaw-RL: Train Any Agent Simply by Talking
- SEED: Self-Evolving On-Policy Distillation for Agentic Reinforcement Learning
- Learning to Reason under Off-Policy Guidance
- Qwen3 Technical Report
- ReAct: Synergizing Reasoning and Acting in Language Models
- $\tau$-bench: A Benchmark for Tool-Agent-User Interaction in Real-World Domains
- OPDSearch+: On-Policy Distillation with RL Refinement for Search-Augmented Reasoning
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks