Blog

Welcome to my blog! Here I share insights from my daily readings of AI research papers and write about interesting open-source projects I’m working on. I hope these posts are helpful to the community and spark interesting discussions.

AI Daily Paper Summary

I read AI and ML research papers daily and post concise summaries here. Each entry covers the key contributions, methods, and takeaways from recent papers across topics like deep learning, reinforcement learning, NLP, computer vision, and more.

Latest — August 19, 2026

1. StateM — Scales the agent harness itself to hit 95.3% raw accuracy on Terminal-Bench 2.1 for a $15 frontier run. arxiv.org/abs/2608.15089
2. Agentic ESOpt — Fine-tunes long-horizon LLM agents via evolution strategies, sidestepping RL’s heavyweight backprop and sparse-reward pain. arxiv.org/abs/2608.17310
3. Demystifying Agent Skills — Isolates when structured skill packages actually help LLM agents at inference time and where they quietly fail. arxiv.org/abs/2608.14036
4. ASI-Bench — Frames a benchmark for artificial superintelligence around creating new, verifiable knowledge rather than compressing existing knowledge. arxiv.org/abs/2608.17271
5. Embodied-Navigator — A point-think-memorize-align framework that aligns VLM navigation with 2D pre-training priors for efficient embodied movement. arxiv.org/abs/2608.17512
6. FreeToken — An edge-native MoE serving system that treats a personal machine as an elastic, bandwidth-adaptive inference platform. arxiv.org/abs/2608.16157
7. EDITBRIDGE — Pushes faithful, efficient image editing past the 1K-resolution wall that quadratic attention normally imposes. arxiv.org/abs/2608.18063
8. HarmProfile — Studies harmful LLM outputs as objects of analysis, characterizing the distributions produced during model misbehavior. arxiv.org/abs/2608.14577
9. Agent Lightning v1.0 — Matures the disaggregated agent-harness-to-RL architecture into a harnessed, production-oriented agentic RL stack. arxiv.org/abs/2608.17528
10. V-RAE — Rethinks video latent spaces to add high-level semantic organization beyond pixel-level reconstruction for generation. arxiv.org/abs/2608.13556

Browse all paper summaries →

Interesting Projects

In this section, I share outcomes, results, and lessons learned from open-source projects I’m working on. The goal is to give back to the community and document the journey of building and experimenting with different ideas.

Browse all project posts →