verl-agent is an open-source implementation of Group-in-Group Policy Optimization (GiGPO), a reinforcement learning algorithm designed to improve credit assignment for long-horizon LLM-based agents. It achieves better performance on benchmarks such as ALFWorld and WebShop compared to baselines while using the same GPU memory and rollout procedures. The project is associated with a research paper and targets developers building or experimenting with autonomous LLM agents.
In the Autonomous agents & workflows space, Paper page takes a focused approach. Inefficient credit assignment and high computational cost when training LLM agents on long-horizon tasks with sparse rewards. It is built as an open-source project for AI researchers and developers. Paper page is open source under the Apache-2.0 license. It runs on the web, and it can be self-hosted.
Behind Paper page is langfengQ, and it first shipped in 2024. The project is developed in the open on GitHub with 2.1k stars and 2 commits in the last 90 days. Among its 4 catalogued features are RL Training, Credit Assignment, and Long-horizon Agents.
Summary written by a language model from the project’s public pages.
What PulseGate has recorded for this listing
Same category — not a similarity match