Yang Zhang
10 papers · Latest:
FlowCompile: An Optimizing Compiler for Structured LLM Workflows
FlowCompile is an optimizing compiler for structured LLM workflows that explores design space at compile-time to find efficient, reusable configurations.
Pop Quiz Attack: Black-box Membership Inference Attacks Against Large Language Models
Introduces PopQuiz, a black-box membership inference attack that turns data into quizzes to reveal if LLMs memorized specific training examples.
Task-Aware Scanning Parameter Configuration for Robotic Inspection Using Vision Language Embeddings and Hyperdimensional Computing
This paper introduces ScanHD, a hyperdimensional computing framework that autonomously configures robotic laser profilers using vision-language embeddings.
Observation of the Magnus Nonlinear Hall effect from Chiral Weyl Monopoles
This paper observes the Magnus Nonlinear Hall effect in CoSi, revealing a new skew-scattering mechanism driven by chiral Weyl monopoles.
PRTS: A Primitive Reasoning and Tasking System via Contrastive Representations
PRTS is a VLA model that uses contrastive Goal-Conditioned RL to learn goal-reachability, significantly improving robot task execution and long-horizon planning.
OCR-Memory: Optical Context Retrieval for Long-Horizon Agent Memory
OCR-Memory enables LLM agents to retain long-term experience by encoding historical trajectories visually, overcoming text-context limits and reducing hallucination.
LLM-ReSum: A Framework for LLM Reflective Summarization through Self-Evaluation
LLM-ReSum is a self-reflective framework that uses LLM-based evaluation to improve summary quality without model finetuning.
TwoHamsters: Benchmarking Multi-Concept Compositional Unsafety in Text-to-Image Models
TwoHamsters benchmarks "Multi-Concept Compositional Unsafety" in T2I models, showing current defenses fail to prevent unsafe content from benign concept combinations.
Seeing but Not Thinking: Routing Distraction in Multimodal Mixture-of-Experts
This paper identifies "Seeing but Not Thinking" in multimodal MoE models, where visual inputs cause routing distraction, and proposes an intervention.
SD-FSMIS: Adapting Stable Diffusion for Few-Shot Medical Image Segmentation
SD-FSMIS adapts Stable Diffusion for few-shot medical image segmentation, achieving competitive results and strong cross-domain generalization.
📬 Weekly AI Paper Digest
Get the top 10 AI/ML arXiv papers from the week — summarized, scored, and delivered to your inbox every Monday.