BunshinRL: Whole-Workspace Forking for Agentic Reinforcement Learning Rollouts
cs.AI, cs.OS
Submitted: 2026-11-07
Updated: 2026-11-11
Code: https://github.com/WukLab/ForkRL
Terminology
Sources
- DeltaBox: Millisecond-Scale Sandbox Checkpoint and Rollback for Stateful AI Agents
- Marginally Correct Tool Caches Can Reverse Group-Normalized Policy Updates
- Reducing Latency of LLM Search Agent via Speculation-based Algorithm-System Co-Design
- Tree Search for LLM Agent Reinforcement Learning
- Kimi K3: Open Frontier Intelligence
- TVCACHE: A Stateful Tool-Value Cache for Post-Training LLM Agents
- Optimizing Agentic Language Model Inference via Speculative Tool Calls
- Seer: Online Context Learning for Fast Synchronous LLM Reinforcement Learning
- Qwen3-VL Technical Report
- DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
- Parallelizing Tool Execution and LLM Generation for Low-Latency Agent Serving
- Crab: A Semantics-Aware Checkpoint/Restore Runtime for Agent Sandboxes
- OSWorld: Benchmarking Multimodal Agents for Open-Ended Tasks in Real Computer Environments
- CUA-Sandbox: Efficient Environments for Computer-Use Agent Reinforcement Learning
- Speculative Actions: A Lossless Framework for Faster Agentic Systems
- ExACT: Teaching AI Agents to Explore with Reflective-MCTS and Exploratory Learning
- ToolCaching: Towards Efficient Caching for LLM Tool-calling
- The Landscape of Agentic Reinforcement Learning for LLMs: A Survey
Related papers
- Multi-Objective Structured Pruning of LLMs for Latency and Model Size Optimization
- MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
- Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
- Embeddings for Preferences, Not Semantics
- Can We Trust Item Response Theory for AI Evaluation?
- Diagnosing Task Insensitivity in Language Agents