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 — November 8, 2026

1. Qwen3-TTS — Combines multilingual voice cloning and description-based voice control with low-latency streaming speech synthesis. arxiv.org/abs/2601.15621
2. Latent Tool Context — Compresses verbose tool observations into latent tokens that preserve an agent’s next actions while reducing context and runtime costs. arxiv.org/abs/2611.06500
3. Signed but Steerable — Shows why signed RAG documents still need checks that bind retrieval vectors to the text they claim to represent. arxiv.org/abs/2611.06498
4. Jailbreak Seed Variability — Finds that identically safety-trained checkpoints fail on different jailbreaks, challenging evaluations based on a single pretraining run. arxiv.org/abs/2610.34074
5. NPCBench — Tests coherent clinical decisions across a patient’s care pathway rather than treating strong isolated medical-task scores as sufficient. arxiv.org/abs/2611.08369
6. UniRestorer++ — Routes image restoration through degradation-aware experts to improve scaling and transfer to unfamiliar real-world corruptions. arxiv.org/abs/2611.06042
7. ViT-HarDis — Encourages attention heads to specialize on difficult face identities without adding inference-time computation. arxiv.org/abs/2610.41789
8. HERIT — Uses retrieval-generated translation labels to broaden access to historical Hanja archives across underrepresented time periods. arxiv.org/abs/2611.04949
9. SynthFaces — Generates balanced synthetic identities for personalized image generation, reducing reliance on real personal photographs. arxiv.org/abs/2610.40119
10. DRAP — Adapts robot-policy learning rates to closed-loop risk so continual learning better retains earlier behaviors without replay. arxiv.org/abs/2611.05029

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 →