memory
Agentic Memory
Continual learning of AI agents — in-context learning, continual fine-tuning, and unlearning.
about
“Context is the new weight.”
I mainly work on post-training LLM agents. Low-latency control of what to remember, forget, and explore decides next-gen world-model-aware, self-improving AI. I work on continual learning, unlearning, memory management, task adaptation — at the agentic context (in-context RL, harness) and model level (distillation, finetuning).
what i work on
world model
World Model
In-context world models, adaptation to post-training task worlds, and adapting agents in evolving envs.
latency
Low-Latency AI
Efficient attention architectures, KV-cache compression, latent segmentation, and recurrent transformers.
safety
AI Safety
Synthetic data training, risks of multi-agent interaction, post-training guardrails, and AI behavioral study.
news & media
- 2026-05· preprintarXiv preprint: SHRED — Document Unlearning via Self-Distillation and Entropy Demotion
- 2026-03· preprintarXiv preprint: Expert Personas Improve LLM Alignment but Damage Accuracy — Bootstrapping Intent-Based Persona Routing with PRISM
- 2026-03· coverageMedia coverage on PRISM paper — The Register, AIToday, Tencent News, 36Kr, QbitAI, Yahoo Tech
- 2025-12· fellowshipWrapped up Project Canary
- 2025-10· talksPresented Multimodal Synthetic Data Finetuning and Model Collapse at ACM ICMI
- 2025-08· fellowshipJoined Project Canary (Handshake AI)
- 2025-07· fellowshipStarted Handshake AI Fellowship 2025
my path
- 201601physics
Physics
photonics & metasurface design · dynamic systems
- 201802biophysics
Agile Systems
bio-inspired flight, sensing, and locomotion
- 202103robotics
Robotics · RL
policy learning for physical control and agent behavior
- 202204vae
Continual Learning · VAE
regularization design for variational autoencoders
- 202305multimodal
Multimodal Generation
diffusion models · vision-language model architecture
- 202506multiagent
Multiagent
coordination, division of labor, and mutual verification across agents — continual adaptation at the population level
- 202607agentic-memory
Agentic Memory
in-context learning, continual fine-tuning, unlearning, and memory scaffolds — adapting agents at the context and model level
- 08horizon
World Models
predictive world models and the architectures to serve them in real time
collaborators
