ComfyUI-Copilot: An Intelligent Assistant for Automated Workflow Development

An LLM-powered plugin designed to simplify and accelerate workflow creation in ComfyUI, an open-source AI art platform, by providing intelligent node/model recommendations and automated workflow generation.

AReaL: A Large-Scale Asynchronous Reinforcement Learning System for Language Reasoning

AReaL is an asynchronous reinforcement learning system that efficiently trains large language models for reasoning tasks by maximizing GPU usage and decoupling generation from training.

Pseudo-Simulation for Autonomous Driving

Pseudo-simulation is a new evaluation paradigm for autonomous vehicles that blends the realism of real-world data with the generalization power of simulation, enabling robust, scalable testing without the need for interactive environments.

SmolVLA: A Vision-Language-Action Model for Affordable and Efficient Robotics

SmolVLA is a compact, open-source VLA model built for low-cost training and real-world deployment on consumer hardware, enabling efficient language-driven robot control without sacrificing performance.

ProRL: Prolonged Reinforcement Learning Expands Reasoning Boundaries in Large Language Models

This paper introduces ProRL, a method that uses long-horizon reinforcement learning to unlock new reasoning strategies in LLMs—strategies that base models cannot access, even with extensive sampling.

The Surprising Effectiveness of Negative Reinforcement in LLM Reasoning

This paper shows that punishing wrong answers—without explicitly rewarding the correct—can be surprisingly effective for improving reasoning in large language models trained via reinforcement learning with verifiable rewards.

CoT is Not True Reasoning, It Is Just a Tight Constraint to Imitate

This paper challenges the idea that Chain-of-Thought (CoT) prompting enables true reasoning in LLMs, arguing instead that CoT acts as a structural constraint that guides models to imitate the appearance of reasoning.

Why Gradients Rapidly Increase Near the End of Training

This note investigates a sudden rise in gradient norms during the late stages of LLM training and identifies a surprising cause: the interplay between weight decay, normalization layers, and scheduled learning rate decay.

General agents need world models

This paper proves that any general agent capable of reliably completing diverse, goal-directed tasks must implicitly learn a predictive model of its environment—challenging the notion that model-free learning is sufficient for general intelligence.

Beyond the 80/20 Rule

This paper investigates how a small subset of high-entropy tokens—termed "forking tokens"—drives the performance of reinforcement learning with verifiable rewards (RLVR) in reasoning tasks for large language models.

- Pseudo-Simulation for Autonomous Driving
- AReaL: A Large-Scale Asynchronous Reinforcement Learning System for Language Reasoning
- ComfyUI-Copilot: An Intelligent Assistant for Automated Workflow Development

- The Surprising Effectiveness of Negative Reinforcement in LLM Reasoning
- ProRL: Prolonged Reinforcement Learning Expands Reasoning Boundaries in Large Language Models
- SmolVLA: A Vision-Language-Action Model for Affordable and Efficient Robotics

- General agents need world models
- Why Gradients Rapidly Increase Near the End of Training
- CoT is Not True Reasoning, It Is Just a Tight Constraint to Imitate: A Theory Perspective

🚨Another surge in progress for reinforcement learning this week, provided by Beyond the 80/20 Rule, ProRL, and AReal all pushing the boundaries.🚀

Check out the top 10 papers for the week👇

- Beyond the 80/20 Rule: High-Entropy Minority Tokens Drive Effective Reinforcement Learning for LLM Reasoning

HiDream-I1

HiDream-I1 is a 17B-parameter open-source image generation model using a novel sparse Diffusion Transformer (DiT) with dynamic Mixture-of-Experts (MoE) to deliver state-of-the-art image quality in seconds while reducing computation.

Silence is Not Consensus

This work introduces the Catfish Agent, a specialized large language model designed to disrupt premature consensus—called Silent Agreement—in multi-agent clinical decision-making systems by injecting structured dissent to improve diagnostic accuracy.

WebDancer: Towards Autonomous Information Seeking Agency

WebDancer is an end-to-end autonomous web agent designed for complex, multi-step information seeking. It combines a data-centric and training-stage pipeline to enable robust reasoning and decision-making in real-world web environments.

LiteCUA

The authors introduce AIOS 1.0, a platform that helps language models better understand and interact with computers by transforming them into contextual environments. Built on this, LiteCUA is a lightweight agent that uses this structured context to perform digital tasks.

Darwin Godel Machine

The Darwin Gödel Machine (DGM) is a self-improving AI system that rewrites its own code to enhance coding performance. Inspired by Gödel machines and Darwinian evolution, it uses empirical validation and an archive of past agents to drive open-ended, recursive improvement.

The Entropy Mechanism of Reinforcement Learning for Reasoning Language Models

This paper analyzes a fundamental barrier in reinforcement learning (RL) for large language models (LLMs): the sharp early collapse of policy entropy, which limits exploration and caps downstream performance.

AgriFM

AgriFM is a multi-source temporal remote sensing foundation model tailored for crop mapping. It introduces a modified Video Swin Transformer backbone for unified spatiotemporal processing of satellite imagery from MODIS, Landsat-8/9, and Sentinel-2.

WorldEval

This paper introduces WorldEval, a real-to-video evaluation framework that uses world models to assess real-world robot manipulation policies in a scalable, safe, and reproducible way. It avoids costly real-world evaluations by simulating robot actions via generated videos.

Learning to Reason without External Rewards

This paper introduces RLIF, a paradigm where LLMs improve reasoning using intrinsic signals instead of external rewards. The authors propose INTUITOR, which uses a model’s self-confidence—measured as self-certainty—as the sole reward signal.

Paper2Poster

The paper introduces Paper2Poster, the first benchmark for automated academic poster generation, and PosterAgent, a visual-in-the-loop multi-agent system that converts research papers into high-quality posters using open-source models.

- WebDancer: Towards Autonomous Information Seeking Agency
- Silence is Not Consensus: Disrupting Agreement Bias in Multi-Agent LLMs via Catfish Agent for Clinical Decision Making
- HiDream-I1: A High-Efficient Image Generative Foundation Model with Sparse Diffusion Transformer

- The Entropy Mechanism of Reinforcement Learning for Reasoning Language Models
- Darwin Godel Machine: Open-Ended Evolution of Self-Improving Agents
- LiteCUA: Computer as MCP Server for Computer-Use Agent on AIOS

- Learning to Reason without External Rewards
- WorldEval: World Model as Real-World Robot Policies Evaluator
- AgriFM: A Multi-source Temporal Remote Sensing Foundation Model for Crop Mapping

🚨There’s a new ceiling for efficient reasoning with the rise of Learning to Reason without External Rewards, along with AgriFM pushing the boundaries of AI to even agriculture🚀

Check out the top 10 papers for the week👇

- Paper2Poster: Towards Multimodal Poster Automation from Scientific Papers

Revealing economic facts: LLMs know more than they say

This paper shows that hidden states (embeddings) of large language models (LLMs) contain rich economic information that can be used to estimate and impute economic and financial statistics more accurately than the LLMs’ text outputs.

LightLab

LightLab presents a diffusion-based method for precise, parametric control over light sources in a single image, enabling users to edit light intensity and color with photorealistic results. The approach fine-tunes a diffusion model on real paired-photo and synthetically rendered data.