Reinforcement learning from verifiable rewards (RLVR) has become a promising paradigm for improving the reasoning capabilities of large language models (LLMs). However, existing RL-based training remains only a rough approximation to human learning, where humans leverage external experience to guide exploration and gradually internalize problem-solving strategies into stable knowledge. Motivated by this gap, we ask: how can LLMs better utilize and internalize experience during RLVR training? To answer this question, we propose Dual Guidance Optimization (DGO), which treats experience as a bridge between external guidance and internal policy learning. Rather than using experience only as an inference-time aid or trajectories only as optimization signals, DGO couples them in a closed loop: experience guides exploration, improved trajectories renew the experience bank, and guided trajectories are rewritten as self-contained traces for policy distillation. This enables the model to better utilize external experience while yielding consistent gains under experience-free inference. Experiments across multiple benchmarks show that DGO consistently outperforms baselines under both experience-free and experience-guided inference, suggesting that coupling utilization and internalization leads to effective reasoning.

Recent advances in large language models (LLMs) have enabled software engineering agents to solve complex code modification tasks. However, most existing approaches rely on execution feedback from Docker-based environments, which require dependency-complete setup, program execution, and test validation, making agent training costly and difficult to scale. We propose SWE-World, a Docker-free framework that replaces containerized execution during downstream agent optimization with a learned surrogate environment. Trained on real agent-environment interaction traces, SWE-World predicts both step-level execution feedback and final test outcomes, preserving the standard agent-environment loop without repeatedly constructing or running Docker environments. This surrogate also enables test-time scaling (TTS) by scoring multiple candidate trajectories through simulated evaluation. Experiments on SWE-bench Verified show that SWE-World improves Qwen2.5-Coder-32B-Instruct from 6.2% to 52.0% with Docker-free SFT, 55.0% with Docker-free RL, and 68.2% with TTS. The code is available at https://anonymous.4open.science/r/SWE-World-2D1E.

Post-training large language models often require optimizing multiple objectives simultaneously, motivating multi-reward reinforcement learning (RL) methods that model different objectives as separate reward signals. Existing methods like Group reward-Decoupled Policy Optimization (GDPO) compute group-normalized advantages separately for each reward dimension and then aggregate them for policy updates. However, this aggregation process can cause cross-reward advantage conflict: the same rollout may have positive advantages on some reward dimensions but negative advantages on others, causing positive and negative signals to cancel each other during aggregation. To address this issue, we propose GD²PO, short for Group-Dynamic Reward-Decoupled Policy Optimization, a conflict-aware multi-reward policy optimization method. Before advantage aggregation, GD²PO uses conflict-aware criteria to identify and filter rollouts with severe reward-wise disagreement, reducing the influence of unreliable update signals. It further introduces query-level reweighting to dynamically adjust the update strength of each query according to its overall reward consensus. Experiments on tool calling and helpfulness-safety alignment show that GD²PO outperforms existing methods. Our code is available at https://github.com/Qwen-Applications/GD2PO.

We propose a topic-guided variational auto-encoder (TGVAE) model for text generation. Distinct from existing variational auto-encoder (VAE) based approaches, which assume a simple Gaussian prior for latent code, our model specifies the prior as a Gaussian mixture model (GMM) parametrized by a neural topic module. Each mixture component corresponds to a latent topic, which provides a guidance to generate sentences under the topic. The neural topic module and the VAE-based neural sequence module in our model are learned jointly. In particular, a sequence of invertible Householder transformations is applied to endow the approximate posterior of the latent code with high flexibility during the model inference. Experimental results show that our TGVAE outperforms its competitors on both unconditional and conditional text generation, which can also generate semantically-meaningful sentences with various topics.