Yan Zhan

YanZhanPKU's activity

• All
• Models
• Datasets
• Spaces
• Buckets
• Papers
• Collections
• Community
• Posts
• Upvotes
• Likes
• Articles

upvoted a paper 4 days ago

Same Bytes, Different Authority: Reserved-Token Representations in Chat-Template Prompt Injection

Paper • 2609.35932 • Published 6 days ago • 7

upvoted a paper 5 days ago

Why Deterministic PRM Guidance Underperforms in Discrete Diffusion Reasoning

Paper • 2609.35472 • Published 6 days ago • 2

upvoted 2 collections 7 days ago

Byte Authority

Collection

Byte Authority code and evaluation records for reserved-token prompt injection. • 2 items • Updated 4 days ago • 1

dLLM PRM Gap

Collection

Adapters and trajectory artifacts for matched-compute PRM guidance and ORM reranking in discrete diffusion reasoning. • 13 items • Updated 5 days ago • 1

upvoted 2 collections 9 days ago

SLCA-GRPO

Collection

Segment-Locked Credit Assignment for tool-calling RL — processed data splits and the reference implementation. • 2 items • Updated 9 days ago • 1

Entropy-Valley

Collection

Training-free length selection for masked diffusion MT (EMNLP 2026). Code: github.com/Entropy-Valley/Entropy-Valley • 5 items • Updated 9 days ago • 3

upvoted a paper 9 days ago

SLCA-GRPO: Resolving Cross-Segment Credit Misattribution in Tool-Calling RL

Paper • 2609.29050 • Published 10 days ago • 13

upvoted a paper 30 days ago

RefineSVG: Visual Feedback-Driven Reinforcement Learning for Image-to-SVG Generation

Paper • 2607.27699 • Published Jul 30 • 1

upvoted 2 papers about 1 month ago

TrustRoboReward: Preference-Ordered Isotonic Score Editing for Multi-Paradigm Robot Reward Models

Paper • 2608.08491 • Published Aug 9 • 1

Length-Adaptive Decoding for Masked Diffusion Machine Translation

Paper • 2608.22274 • Published Aug 23 • 4

upvoted 2 papers about 1 year ago

TrustJudge: Inconsistencies of LLM-as-a-Judge and How to Alleviate Them

Paper • 2509.21117 • Published Sep 25, 2025 • 30

Temporal Self-Rewarding Language Models: Decoupling Chosen-Rejected via Past-Future

Paper • 2508.06026 • Published Aug 8, 2025 • 15