PhD Student @ZuckermanBrain and @EE_ColumbiaSEAS. Interested in Human&Machine Intelligence
NYC
Joined October 2025
- Tweets30
- Following174
- Followers49
- Likes5.2K
Linyang He retweeted
Congratulations to the @NimaMesgarani lab’s Xiaomin He and alumnus @infinivishal, who received the Best Student Paper Award at #interspeech2026 for their work on brain-controlled hearing research. 👏🧠👏
Read more: isca-speech.org/ISCA-Awards
@ColumbiaMed
Linyang He retweeted
1/3 🎉 Excited to announce BilingLM 2027!
The Workshop on Bilingualism and Language Modeling will be held at #COLING2027 in Macau! 🇲🇴
Join us to explore bilingual learning, representation, and use across humans and machines.
🌐 bilinglm.github.io/
Linyang He retweeted
Late to the party, but happy to introduce our #ECCV2026 paper: On the Diffusibility of High-Dimensional Latents.
We found that finetuning representation autoencoders (RAEs) for reconstruction can hurt diffusibility under standard velocity prediction in flow matching. Why?(1/10)
Linyang He retweeted
Introducing FlashForward: A faster autoregressive diffusion generation pipeline that delivers consistent and higher-quality videos, tested on videos up to 65 seconds long.
PJ page: yikai-wang.github.io/FlashFo…
arxiv: arxiv.org/abs/2609.32540
HF paper page: huggingface.co/papers/2609.3…
Linyang He retweeted
I'm excited to introduce Prove2Me (prove2.me/), an open, collaborative, agent-native platform for scaling the formalization of mathematics. Prove2Me was recently the collaboration platform behind Anthropic's formalization of Fermat's Last Theorem (anthropic.com/research/forma…). This is the first complete computer-checked proof of the theorem. Claude agents working together through Prove2Me wrote 13 million lines of Lean in 11 days.
Our mission is to formalize every research paper, past and future. Formalized papers make peer review faster and more trustworthy, and they give everyone a single verified foundation that any human or agent can build on. We are equally committed to making formalized mathematics easier for people to explore and question, because we believe human understanding is not something AI can replace. We would love to build this together with every research community that uses mathematics, and we hope Prove2Me becomes a tool that benefits everyone.
Two design choices sit at the core of Prove2Me: a directed acyclic graph (DAG) of theorem statements, and an agent-native API protocol (arxiv.org/abs/2608.28433). Together they let any agent connect to the platform and contribute formalizations to one ever-growing graph that may eventually contain all of mathematics. Today the platform holds 22.3 million lines of Lean and 70,000 theorems, and we can't wait to see it grow.
Everyone is welcome to contribute, whether you are an expert or a hobbyist. The easiest way to start is to open a coding agent (Codex, Claude Code, or any other) and say "go to prove2.me and find something interesting to work on", or "please help me formalize this paper" with an arXiv link. Any feedback and comments will be welcome!
Linyang He retweeted
Through the VOICE trial, Terry is using his Neuralink implant to help fine-tune a brain-to-voice interface for himself and others who can’t speak. He trained the algorithm first by miming speech as best he could, then by simply thinking the words and hearing them come out in his own natural voice.
Powered by Grok Voice from @spacexai
Linyang He retweeted
Releasing the model weights and technical report of Kimi K3.
Kimi K3 is our most capable model: a 2.8T MoE model with native visual understanding and a 1M-token context window.
New model architecture: 2.5x the intelligence per unit of compute, not just more params.
Alongside Kimi K3, we're opening up more of the stack behind it — high-performance attention kernels, MoE communication library, and infrastructure for running agent environments at scale.
Model weights: huggingface.co/moonshotai/Ki…
Tech report: github.com/MoonshotAI/Kimi-K…
Tech blog: kimi.com/blog/kimi-k3
Linyang He retweeted
“More on-policy” is NOT always better!
ReOPD reveals a two-sided prefix shift:
① student occupancy mismatch
② teacher reliability mismatch
Importance sampling builds a reliability-aware bridge over offline prefixes.
0 tool calls; ≥4× faster.
arxiv.org/abs/2607.04763
Linyang He retweeted
LiveMathematicianBench has been accepted to COLM 2026.
It is a live benchmark built around newly published mathematical results, making it harder for models to rely on memorization and better suited for testing research-level reasoning.
Between theorem recognition and theorem proving lies theorem understanding.
We introduce LiveMathematicianBench: a live, contamination-resistant testbed for research-level mathematical reasoning, built from post-cutoff arXiv theorems.
It probes a capability that existing benchmarks rarely isolate: whether models can understand theorem statements, track delicate assumptions, reason over logical structure, and leverage proof-level guidance.
livemathematicianbench.githu…
Linyang He retweeted
How does the brain combine what we see with what language tells us to expect? Excited to share two recent works from our group and collaborators, connecting vision, language, and multimodal AI. 🧵1/11
Linyang He retweeted
🚀 Introducing SkillOpt — an optimizer for agent skills.
Instead of finetuning model weights, we treat a natural-language skill as a trainable external parameter.
Think of it as deep learning for the frontier-model + agent era: learning rate, LR schedule, mini-batch, batch size, epoch, momentum — all in text-space optimization.
SkillOpt enables stable, controllable skill updates through bounded edits, allowing the optimizer to summarize “gradient directions” from agent experience and continuously improve procedural capability.
We evaluate SkillOpt across 6 benchmarks and 7 models, under both direct model calls and real agent execution loops with Codex + Claude Code. SkillOpt achieves best or tied-best results in 52/52 settings.
Train the skill, not the model. 🛠️🤖
🌐 aka.ms/skillopt
📄 huggingface.co/papers/2605.2…
Linyang He retweeted
Hearing aids amplify all incoming sound, and so struggle with noisy surroundings. Brain-controlled hearing tech from @NimaMesgarani, @infinivishal & team could lead to a new generation of hearing systems that help people single out a voice in a crowd.
Read more: zuckermaninstitute.columbia.…
Linyang He retweeted
🚀 Announcing the Chinese BabyLM Challenge: the first shared task on data-efficient pretraining for Chinese.
📍 Co-located with NLPCC 2026 (Nov 3–5, Macau🇨🇳🇲🇴)
Can you train a strong Chinese LM on just ~100M words? chinese-babylm.github.io
🧵 👇(1/6)
Linyang He retweeted
1/7 How can we develop decoding methods that generalize to novel subjects without any fine-tuning?
Excited to share our recent collaboration to appear at CVPR 2026!
Work led by Andrew Luo's group, with many collaborators (@_jacobprince_ , Mike Tarr, @KriegeskorteLab and others).
code: github.com/ezacngm/brainCode…
paper: arxiv.org/abs/2604.08537
Linyang He retweeted
I think we finally made really significant progress on the biggest unsolved "developmental AI" problem: learning from human-scale data. Key idea: zero-shot world models that support concept extraction via approximate causal inference. amazing collab w/ @khai_loong_aw @mcxfrank
Today's best AI needs orders of magnitude more data than a human child to achieve visual competence.
We introduce the Zero-shot World Model (ZWM), an approach that substantially narrows this gap. Even when trained on the first-person experience of a single child, BabyZWM matches state-of-the-art models on diverse visual-cognitive tasks – with no task-specific training, i.e., zero-shot. 🧵
Linyang He retweeted
Between theorem recognition and theorem proving lies theorem understanding.
We introduce LiveMathematicianBench: a live, contamination-resistant testbed for research-level mathematical reasoning, built from post-cutoff arXiv theorems.
It probes a capability that existing benchmarks rarely isolate: whether models can understand theorem statements, track delicate assumptions, reason over logical structure, and leverage proof-level guidance.
livemathematicianbench.githu…
Linyang He retweeted
SFT curates responses. RL curates sampling.
RL improves by curating what the model experiences: condition; distribution; weighting of what gets learned from.
Better signal curation shifts the performance-compute curve upward.
Full write-up below 👇
hendrydong.github.io/blogs/p…
Linyang He retweeted
#VideoReason We are open-sourcing the entire VBVR stack to speed-up the arrival of video reasoning as the next fundamental paradigm of intelligence
- 150+ synthetic generators
- 1 million training clips
- Cloud-scale data factory
- Unified EvalKit
- 100 rule-based evaluators
- Strong baseline model
Checkout at video-reason.com
