Template collapse is a distinct failure mode in agentic RL invisible to entropy; mutual information proxies diagnose it better and SNR-aware filtering using reward variance improves input-dependent reasoning and task performance across planning, math, navigation, and code tasks.
High-dimensional continuous control using generalized advantage estimation
5 Pith papers cite this work. Polarity classification is still indexing.
5
Pith papers citing it
representative citing papers
MemAgent uses multi-conversation RL to train a memory agent that reads text in segments and overwrites memory, extrapolating from 8K training to 3.5M token QA with under 5% loss and 95%+ on 512K RULER.
DAPO introduces decoupled clipping and dynamic sampling for LLM RL, achieving 50 on AIME 2024 with Qwen2.5-32B while fully open-sourcing code, data, and the verl-based training system.
REINFORCE-style variants outperform PPO, DPO, and RAFT in RLHF for LLMs by removing unnecessary PPO components and adapting the simpler method to LLM alignment characteristics.
HAPO is a new token-level policy optimization method for LLMs that continuously adapts four optimization stages using entropy, claiming consistent gains over DAPO on math, code, and logic tasks.
citing papers explorer
-
RAGEN-2: Reasoning Collapse in Agentic RL
Template collapse is a distinct failure mode in agentic RL invisible to entropy; mutual information proxies diagnose it better and SNR-aware filtering using reward variance improves input-dependent reasoning and task performance across planning, math, navigation, and code tasks.
-
MemAgent: Reshaping Long-Context LLM with Multi-Conv RL-based Memory Agent
MemAgent uses multi-conversation RL to train a memory agent that reads text in segments and overwrites memory, extrapolating from 8K training to 3.5M token QA with under 5% loss and 95%+ on 512K RULER.
-
DAPO: An Open-Source LLM Reinforcement Learning System at Scale
DAPO introduces decoupled clipping and dynamic sampling for LLM RL, achieving 50 on AIME 2024 with Qwen2.5-32B while fully open-sourcing code, data, and the verl-based training system.
-
Back to Basics: Revisiting REINFORCE Style Optimization for Learning from Human Feedback in LLMs
REINFORCE-style variants outperform PPO, DPO, and RAFT in RLHF for LLMs by removing unnecessary PPO components and adapting the simpler method to LLM alignment characteristics.
-
Heterogeneous Adaptive Policy Optimization: Tailoring Optimization to Every Token's Nature
HAPO is a new token-level policy optimization method for LLMs that continuously adapts four optimization stages using entropy, claiming consistent gains over DAPO on math, code, and logic tasks.