Jun Wang
6 papers · Latest:
FollowTable: A Benchmark for Instruction-Following Table Retrieval
FollowTable introduces a new benchmark and metric for Instruction-Following Table Retrieval (IFTR), revealing existing models struggle with fine-grained instructions.
TingIS: Real-time Risk Event Discovery from Noisy Customer Incidents at Enterprise Scale
TingIS is an enterprise-scale system using LLMs and noise reduction to discover real-time risk events from noisy customer incidents with high accuracy.
Discrete Preference Learning for Personalized Multimodal Generation
DPPMG learns discrete modal-specific preferences to generate personalized and consistent multimodal content from user interactions.
R3D: Revisiting 3D Policy Learning
R3D introduces a stable 3D policy learning architecture with a transformer encoder and diffusion decoder, overcoming overfitting and instability.
Compiling Activation Steering into Weights via Null-Space Constraints for Stealthy Backdoors
This paper introduces a method to inject stealthy and reliable backdoors into LLMs by compiling activation steering vectors into model weights via null-space constraints.
Externalization in LLM Agents: A Unified Review of Memory, Skills, Protocols and Harness Engineering
LLM agents increasingly externalize capabilities like memory, skills, and protocols into surrounding infrastructure, transforming how they solve complex tasks.
📬 Weekly AI Paper Digest
Get the top 10 AI/ML arXiv papers from the week — summarized, scored, and delivered to your inbox every Monday.