Super excited to share Test-Time Matching (TTM), an iterative, self-improving algorithm that unlocks substantial compositional reasoning capabilities in multimodal models.

TTM enables SigLIP-B16 (~0.2B params) to outperform GPT-4.1 on MMVP-VLM, establishing a new SOTA.

🚀Excited to share our new paper:

Online Finetuning Decision Transformers with Pure RL Gradients

RL drives reasoning in LLMs—but remains underexplored for online finetuning of Decision Transformers (DTs), where most methods still rely mainly on supervised objectives.

Why?

Sharing new paper: Towards Multimodal Active Learning: Efficient Learning with Limited Paired Data

We extend classical unimodal active learning to the multimodal AL with unaligned data, allowing data-efficient finetuning and pretraining of vision-language models such as CLIP and SigLIP.

1/3

🚀Excited to share our new paper: Strategic Scaling of Test-Time Compute: A Bandit Learning Approach.

We turn test-time compute allocation into a bandit learning problem, achieving:
✅ +11.10% on MATH-500
✅ +7.41% on LiveCodeBench

Paper: arxiv.org/pdf/2506.12721

(1/3)

There is a ton of interest in the question of whether AI can be be funny: www.bbc.com/future/artic.... Our paper at NeurIPS investigates the humor generation capabilities of the latest and greatest AI models using one of world’s largest humor datasets! arxiv.org/pdf/2406.10522

I’m recruiting multiple PhD students for Fall 2025 at UCR! If you’re interested in working on efficient ML, RL, and LLMs, please apply to the UCR CS/EE PhD program.

Please visit yinglunz.com for detailed information on research directions and contact instructions.