arxiv arXiv cs.CL · just now · research

VHD-Play generates agentic RL environments from solved mechanisms

from English

VHD-Play is a pipeline that generates diverse agentic reinforcement learning environments by sampling and solving mathematical models before rendering their decision processes as stateful tools. This approach ensures that executable dynamics and trajectory-scoring references are inherited directly from the solved model, addressing the misalignment issues common in existing generation pipelines.

The pipeline produces 3,300 diverse agentic environments at a cost of a few cents each.
Training Qwen3.6-35B-A3B on three families raises its mean agentic score from 0.204 to 0.815 in a five-family diagnostic.
Gains extend to held-out instances, eight unseen mechanism families, and external benchmarks for general function calling, travel planning, and e-commerce.
Comparisons between written-out problems and stateful versions show that the learnable gap lies primarily in stateful interaction rather than underlying problem solving.

The authors consider this significant because a frozen 35B setter can realize larger environments, and scale-matched training retains gains as mechanism size and horizon grow, indicating the potential for an evolving training substrate.

Importance 2/3 New harness with differentiators arXiv cs.CL Alibaba (Qwen) AI agents Training methods

Related articles

arxiv arXiv cs.CL · 1h ago · 1 view Live

SkillGym internalizes human skills into LLMs for real-world problem solving

Researchers introduce SkillGym, a framework that transforms human-written agent skills into executable, verifiable training environments for large language model agents. The system utilizes a skill-to-task pipeline to instantiate concrete tasks and verify outcomes with code-based checkers.

arxiv arXiv cs.CL · 15d ago · 25 views

ExecCritic uses role-specific RL to improve coding agents via test-guided repair

Researchers introduce ExecCritic, a framework that combines a test-verify-revise scaffold with role-specific reinforcement learning to enhance coding agents. The system separates test construction from source-code repair, using a Test agent to generate repository-native tests and a Repair agent to revise code based on execution feedback.

arxiv arXiv cs.AI · 98d ago · 7 views

Spotlight: Using Spot GPUs to Accelerate DiT RL Post-Training

Spotlight enables DiT RL post-training by leveraging idle spot GPUs, reducing costs by 1.4-6.4× while achieving superior image quality. It uses stale model weights in exploration and reconfigures sequence parallelism in real time, allowing efficient GPU utilization without breaking training pipelines.

arxiv arXiv cs.LG · 99d ago · 2 views

EnvRL: Leveraging Environment Dynamics in Agentic RL

EnvRL introduces a framework that enhances agentic reinforcement learning by incorporating environment dynamics through state prediction and inverse dynamics objectives. When trained with GRPO, EnvRL improves success rates of Qwen-2.5-1.5B-Instruct from 72.8% to 77.4% on ALFWorld and from 56.8% to 67.0% on WebShop.

arxiv arXiv cs.CL · 99d ago · 1 view

EnvRL: Leveraging Environment Dynamics in Agentic RL

EnvRL introduces a framework that enhances agentic reinforcement learning by incorporating environment dynamics through state prediction and inverse dynamics objectives. It achieves significant gains in success rates on long-horizon benchmarks, improving Qwen-2.5-1.5B-Instruct performance from 72.8% to 77.4% on ALFWorld and from 56.8% to 67.0% on WebShop when trained with GRPO.