Vision-OPD is a regional-to-global on-policy self-distillation framework that transfers a model's own privileged crop-conditioned perception to its full-image policy, enabling fine-grained visual u…

This repo implements the unification of diffusion model and recified flow through variational bridge.

World model reasoning RL for multi-turn VLM agents

SkillRL: Evolving Agents via Recursive Skill-Augmented Reinforcement Learning

[ICML 2026] XSkill: Continual Learning from Experience and Skills in Multimodal Agents

The codebase of Cola DLM

A curated collection of papers and resources on On-Policy Distillation for Large Language Models.

[CVPR2026] Official codebase for the paper "Reasoning Within the Mind: Dynamic Multimodal Interleaving in Latent Space"

Pretraining and inference code for a large-scale depth-recurrent language model

PyTorch-based open-source code for paper "SOD: Step-wise On-policy Distillation for Small Language Model Agents"

On Policy Distillation Build on top of Verl

Repository hosting code for "Actions Speak Louder than Words: Trillion-Parameter Sequential Transducers for Generative Recommendations" (https://arxiv.org/abs/2402.17152).

Minimal reproduction of OneRec

We propose the first Tri-party LLM-agent Recommendation framework (TriRec) that explicitly coordinates user utility, item exposure, and platform-level fairness.

[ACL 2025] iAgent: LLM Agent as a Shield between User and Recommender Systems

2026TAAC腾讯广告算法大赛-KDDCUP方案，best score：0.832321，rank：51

参考 [参赛队伍] TAAC 2026 腾讯广告算法大赛 X KDD 2026

Fixing GRPO training collapse in long-horizon multi-tool agents. A lightweight PRM-Lite + LATA joint approach achieves +37% over vanilla GRPO on τ-bench airline (50-task, multi-turn).

[ACM-RecSys26] Baseline code for music-crs challenge