Zimin Zhang

UIUC CS Master Student

I’m Zimin Zhang, a CS Master student at University of Illinois Urbana-Champaign (UIUC) starting from 2024 Fall, working with Prof. Hao Peng. Previously, I obtained my B.S. degree in Math of Computation and Statistics from University of California, Los Angeles (UCLA), where I was fortunate to work with Prof. Yizhou Sun.

My research focues on LLM post-training and mechanistic interpretability, especially on understanding the internal mechanisms of LLM reasoning and improving LLM reasoning with latent space reasoning using RL.

Selected Publications

The Unreasonable Effectiveness of Entropy Minimization in LLM Reasoning

Shivam Agarwal, Zimin Zhang, Lifan Yuan, Jiawei Han, and Hao Peng

In The Thirty-ninth Annual Conference on Neural Information Processing Systems (NeurIPS), 2025

Abs PDF Code

Entropy minimization (EM) trains the model to concentrate even more probability mass on its most confident outputs. We show that this simple objective alone, without any labeled data, can substantially improve large language models’ (LLMs) performance on challenging math, physics, and coding tasks. We explore three approaches: (1) EM-FT minimizes token-level entropy similarly to instruction finetuning, but on unlabeled outputs drawn from the model; (2) EM-RL: reinforcement learning with negative entropy as the only reward to maximize; (3) EM-INF: inference-time logit adjustment to reduce entropy without any training data or parameter updates. On Qwen-7B, EM-RL, without any labeled data, achieves comparable or better performance than strong RL baselines such as GRPO and RLOO that are trained on 60K labeled examples. Furthermore, EM-INF enables Qwen-32B to match or exceed the performance of proprietary models like GPT-4o, Claude 3 Opus, and Gemini 1.5 Pro on the challenging SciCode benchmark, while being 3x more efficient than self-consistency and sequential refinement. Our findings reveal that many pretrained LLMs possess previously underappreciated reasoning capabilities that can be effectively elicited through entropy minimization alone, without any labeled data or even any parameter updates.
Automated Molecular Concept Generation and Labeling with Large Language Models

Zimin Zhang, Qianli Wu, Botao Xia, Fang Sun, Ziniu Hu, Yizhou Sun, and Shichang Zhang

In Proceedings of the 31st International Conference on Computational Linguistics (COLING), 2025

Abs PDF Code

Artificial intelligence (AI) is transforming scientific research, with explainable AI methods like concept-based models (CMs) showing promise for new discoveries. However, in molecular science, CMs are less common than black-box models like Graph Neural Networks (GNNs), due to their need for predefined concepts and manual labeling. This paper introduces the Automated Molecular Concept (AutoMolCo) framework, which leverages Large Language Models (LLMs) to automatically generate and label predictive molecular concepts. Through iterative concept refinement, AutoMolCo enables simple linear models to outperform GNNs and LLM in-context learning on several benchmarks. The framework operates without human knowledge input, overcoming limitations of existing CMs while maintaining explainability and allowing easy intervention. Experiments on MoleculeNet and High-Throughput Experimentation (HTE) datasets demonstrate that AutoMolCo-induced explainable CMs are beneficial for molecular science research.