Code for kai0, including training, inference and data collection.

Official style files for papers submitted to venues of the Association for Computational Linguistics

Official Repo of "Flow-OPD: On-Policy Distillation for Flow Matching Models"

[ACM MM 2026] On the Limitations of Cross-Lingual Consistency in Multilingual Text-to-Image Generation

Explain in Your Own Words: Improving Reasoning via Token-Selective Dual Knowledge Distillation, The Fourteenth International Conference on Learning Representations (ICLR) 2026, Accepted

OPRD: On-Policy Representation Distillation (https://arxiv.org/abs/2606.06021)

openpi-RLT is an openpi-based real-robot RL system with RL-token-guided action refinement.

[RSS 2026] Code for RISE: Self-Improving Robot Policy with Compositional World Model

Codex-native Academic Research Skills suite for human-in-the-loop academic research workflows

Academic Research Skills for Claude Code: research → write → review → revise → finalize

[ECCV 2026] EgoSim: Egocentric World Simulator for Embodiment Interaction Generation

NEO Series: Native Vision-Language Models from First Principles

The implementation of Teachability-Aware On-Policy Distillation (TA-OPD), the method introduced in Not All Disagreement Is Learnable: Token Teachability in On-Policy Distillation.

Everything about the SmolLM and SmolVLM family of models

On Policy Distillation Build on top of Verl

Rethinking On-Policy Distillation of Large Language Models: Phenomenology, Mechanism, and Recipe

Code, data and weights for the paper **What drives success in physical planning with Joint-Embedding Predictive World Models?**

Official repository for the paper "Learning beyond Teacher: Generalized On-Policy Distillation with Reward Extrapolation"

Post-training with Tinker

HY-Embodied: Embodied Foundation Models for Real-World Agents

Awesome List for On-Policy Distillation

Official implementation of "OneTwoVLA: A Unified Vision-Language-Action Model with Adaptive Reasoning"

Source code for the Refined Policy Distillation paper.

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…

A high-throughput and memory-efficient inference and serving engine for LLMs