#rl

10 Sept 2026 · Paper summary
Paper summary: Scaling Automatic Research Agents via World Models
RL for automatic-research agents is bottlenecked by slow, sandboxed experiment execution; World Model RL replaces execution with a learned world model plus debiasing and denoising, training 3–4× faster.
17 Mar 2026 · Paper summary
Paper summary: AI Can Learn Scientific Taste
Reinforcement Learning from Community Feedback trains a "Scientific Judge" on signals like citations, and a "Scientific Thinker" that proposes research ideas; the judge generalises to future years and unseen fields.
12 Mar 2026 · Paper summary
Paper summary: OpenClaw-RL: Train Any Agent Simply by Talking
OpenClaw-RL turns the next state after every agent action (a user's reply, a tool output, a GUI change) into live training signal, so a personal agent improves online simply by being used.
18 Feb 2026 · Paper summary
Paper summary: GLM-5: from Vibe Coding to Agentic Engineering
GLM-5 targets "agentic engineering" beyond vibe coding, using DSA attention to cut long-context cost and an asynchronous RL infrastructure plus new agent RL algorithms for long-horizon tasks.
3 Feb 2026 · Paper summary
Paper summary: Green-VLA: Staged Vision-Language-Action Model for Generalist Robots
Green-VLA is a five-stage vision-language-action training recipe, from a base VLM to RL policy alignment, with one embodiment-aware action interface that controls humanoids, mobile manipulators and fixed arms.
9 Jan 2026 · Paper summary
Paper summary: GDPO: Group reward-Decoupled Normalization Policy Optimization for Multi-reward RL Optimization
With several rewards (correctness, format, length…), GRPO's group normalisation can collapse different reward combinations into identical advantages; GDPO normalises each reward separately and trains more stably.
14 Oct 2025 · Paper summary
Paper summary: QeRL: Beyond Efficiency -- Quantization-enhanced Reinforcement Learning for LLMs
QeRL combines NVFP4 quantisation with LoRA for RL training of LLMs; beyond saving memory and speeding up rollouts, quantisation noise raises policy entropy and helps exploration. It enables RL on a 32B model with one H100.
19 Sept 2025 · Paper summary
Paper summary: FlowRL: Matching Reward Distributions for LLM Reasoning
FlowRL replaces reward maximisation with matching the whole reward distribution (via a learnable partition function and reverse KL), to keep diverse reasoning paths alive; it reports +10.0% over GRPO on maths.
10 Sept 2025 · Paper summary
Paper summary: Sharing is Caring: Efficient LM Post-Training with Collective RL Experience Sharing
SAPO is a fully decentralised, asynchronous RL post-training method in which many heterogeneous nodes train their own models but share rollouts, reporting up to 94% higher cumulative reward in controlled tests.
29 Jul 2025 · Paper summary
Paper summary: Agentic Reinforced Policy Optimization
ARPO is an RL algorithm for multi-turn tool-using agents that samples extra rollouts right after tool calls, where the model's token entropy spikes, and is reported to beat trajectory-level RL with half the tool calls.
25 Jul 2025 · Paper summary
Paper summary: Group Sequence Policy Optimization
Group Sequence Policy Optimization computes importance ratios and clipping at the level of whole sequences rather than tokens, and is reported to be more stable and efficient than GRPO, especially for MoE models.
2 Jul 2025 · Paper summary
Paper summary: GLM-4.5V and GLM-4.1V-Thinking: Towards Versatile Multimodal Reasoning with Scalable Reinforcement Learning
GLM-4.1V-Thinking and GLM-4.5V use a reasoning-centric training framework with Reinforcement Learning with Curriculum Sampling (RLCS); GLM-4.5V is reported to lead open models of its size on nearly all of 42 benchmarks.
17 Jun 2025 · Paper summary
Paper summary: MiniMax-M1: Scaling Test-Time Compute Efficiently with Lightning Attention
MiniMax-M1 is an open-weight hybrid-attention reasoning model (456B total, 45.9B active, 1M-token context) trained with large-scale RL and a new algorithm, CISPO, at a reported rental cost of $534,700.
10 Jun 2025 · Paper summary
Paper summary: Reinforcement Pre-Training
Reinforcement Pre-Training reframes next-token prediction as a reasoning task: the model reasons before predicting the next token and gets a verifiable reward when it's right, so ordinary text becomes RL training data.
4 Jun 2025 · Paper summary
Paper summary: Reflect, Retry, Reward: Self-Improving LLMs via Reinforcement Learning
When a model fails a task, it writes a self-reflection and tries again; if the retry succeeds, the reflection tokens are rewarded, which trains better self-reflection with only binary feedback.
7 May 2025 · Paper summary
Paper summary: Absolute Zero: Reinforced Self-play Reasoning with Zero Data
Absolute Zero trains a model to propose its own coding and maths-style tasks and then solve them, with a code executor checking both, so reasoning improves through self-play with no external training data.
30 Apr 2025 · Paper summary
Paper summary: Reinforcement Learning for Reasoning in Large Language Models with One Training Example
RL with verifiable rewards on a single maths training example lifts Qwen2.5-Math-1.5B on MATH500 from 36.0% to 73.6%, matching training on a 1.2k-example dataset.
21 Apr 2025 · Paper summary
Paper summary: Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model?
Measuring pass@k at large k, this study finds RL with verifiable rewards makes models better at k = 1 but doesn't expand what the base model can solve at all; distillation, by contrast, can.
29 Jan 2025 · Paper summary
Paper summary: SFT Memorizes, RL Generalizes: A Comparative Study of Foundation Model Post-training
On a card-arithmetic game and a real-world navigation task, models post-trained with RL generalise to unseen rule and visual variants, while SFT tends to memorise the training data.
23 Jan 2025 · Paper summary
Paper summary: DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
DeepSeek-R1 shows that large-scale reinforcement learning, without human-labelled reasoning traces, can bring out advanced reasoning such as self-reflection and verification in a pretrained LLM.
6 Feb 2024 · Paper summary
Paper summary: DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
DeepSeekMath 7B continues pre-training on 120B maths tokens mined from Common Crawl and introduces GRPO, a PPO variant without a value network, reaching 51.7% on the MATH benchmark.