15 citations · 24 across the 14 of their papers we have counts for
16 papers
Are We Recovering Mechanisms? Objective-Level Recovery Gaps in Mechanistic Interpretability
Chuqin Geng, Li Zhang, Haolin Ye +3
Mechanistic interpretability aims to recover the internal computations responsible for model behavior. Progress in automated circuit discovery is often framed as a search problem:…
Rethinking Circuit Evaluation: Do Circuits Explain Model Errors?
Li Zhang, Chuqin Geng, Mark Zhang +4
Mechanistic interpretability (MI) aims to explain a model's behaviour through analyzing its internal computations; circuit-based explanations aim to isolate these computations with…
DreamProver: Evolving Transferable Lemma Libraries via a Wake-Sleep Theorem-Proving Agent
Youyuan Zhang, Jialiang Sun, Hangrui Bi +4
We introduce DreamProver, an agentic framework that leverages a "wake-sleep" program induction paradigm to discover reusable lemmas for formal theorem proving. Existing approaches…
Neural Proposals, Symbolic Guarantees: Neuro-Symbolic Graph Generative Modeling
Chuqin Geng, Li Zhang, Mark Zhang +4
While deep generative models excel at capturing graph data distributions, they struggle to satisfy complex, hard constraints. In unconstrained settings, these models typically prod…
Beyond Message Passing: A Symbolic Alternative for Expressive and Interpretable Graph Learning
Chuqin Geng, Li Zhang, Haolin Ye +5
Graph Neural Networks (GNNs) have become essential in high-stakes domains such as drug discovery, yet their black-box nature remains a significant barrier to trustworthiness. While…
LogicXGNN: Grounded Logical Rules for Explaining Graph Neural Networks
Chuqin Geng, Ziyu Zhao, Zhaoyue Wang +3
Existing rule-based explanations for Graph Neural Networks (GNNs) provide global interpretability but often optimize and assess fidelity in an intermediate, uninterpretable concept…
VisionLogic: Discovering and Grounding Decision-Relevant Visual Concepts
Chuqin Geng, Yuhe Jiang, Ziyu Zhao +7
Concept-based explanations help users understand vision models through recognizable visual patterns. However, existing methods often rely on correlational signals without directly…
NEUROLOGIC: From Neural Representations to Interpretable Logic Rules
Chuqin Geng, Anqi Xing, Li Zhang +3
Rule-based explanation methods offer rigorous and globally interpretable insights into neural network behavior. However, existing approaches are mostly limited to small fully conne…
SAT-DIFF: A Tree Diffing Framework Using SAT Solving
Chuqin Geng, Haolin Ye, Yihan Zhang +2
Computing differences between tree-structured data is a critical but challenging problem in software analysis. In this paper, we propose a novel tree diffing approach called SatDif…
Learning Minimal Neural Specifications
Chuqin Geng, Zhaoyue Wang, Haolin Ye +1
Formal verification is only as good as the specification of a system, which is also true for neural network verification. Existing specifications follow the paradigm of data as spe…
TorchProbe: Fuzzing Dynamic Deep Learning Compilers
Qidong Su, Chuqin Geng, Gennady Pekhimenko +1
Static and dynamic computational graphs represent two distinct approaches to constructing deep learning frameworks. The former prioritizes compiler-based optimizations, while the l…
Can ChatGPT Pass An Introductory Level Functional Language Programming Course?
Chuqin Geng, Yihan Zhang, Brigitte Pientka +1
The recent introduction of ChatGPT has drawn significant attention from both industry and academia due to its impressive capabilities in solving a diverse range of tasks, including…
Identifying Different Student Clusters in Functional Programming Assignments: From Quick Learners to Struggling Students
Chuqin Geng, Wenwen Xu, Yingjie Xu +2
Instructors and students alike are often focused on the grade in programming assignments as a key measure of how well a student is mastering the material and whether a student is s…
Scalar Invariant Networks with Zero Bias
Chuqin Geng, Xiaojie Xu, Haolin Ye +1
Just like weights, bias terms are the learnable parameters of many popular machine learning models, including neural networks. Biases are thought to enhance the representational po…
Novice Type Error Diagnosis with Natural Language Models
Chuqin Geng, Haolin Ye, Yixuan Li +3
Strong static type systems help programmers eliminate many errors without much burden of supplying type annotations. However, this flexibility makes it highly non-trivial to diagno…
Towards Reliable Neural Specifications
Chuqin Geng, Nham Le, Xiaojie Xu +3
Having reliable specifications is an unavoidable challenge in achieving verifiable correctness, robustness, and interpretability of AI systems. Existing specifications for neural n…