-
Unmasking Order in Diffusion Language Models
🧩How diffusion language models choose what to unmask next
-
RAG Knowledge Conflicts
📚When should a RAG model follow or resist retrieved evidence?
-
Query Expansion for Dense Retrieval
📖Compare LLM query expansion and joint retriever training
-
Long-Context Preference Optimization
📖Explore bandit chunk sampling for long-context DPO
Pengcheng Huang
chengpingan
·
AI & ML interests
AI & Agent
Recent Activity
updated a collection about 5 hours ago
Interactive Research Guides updated a collection about 5 hours ago
Interactive Research Guides updated a collection about 5 hours ago
Interactive Research GuidesOrganizations
None yet
Interactive research guides to diffusion decoding, RAG, dense retrieval, long-context DPO, tool use, translation and edge-cloud agents.
Selected Research Papers — Pengcheng Huang
Papers on large language models, including diffusion decoding, retrieval-augmented generation, tool use, and edge–cloud agents.
-
Empirical Analysis of Decoding Biases in Masked Diffusion Models
Paper • 2508.13021 • Published -
PIP-KAG: Mitigating Knowledge Conflicts in Knowledge-Augmented Generation via Parametric Pruning
Paper • 2502.15543 • Published • 2 -
SPARK: Adaptive Parametric Knowledge Participation for Faithful Retrieval-Augmented Generation
Paper • 2606.27786 • Published -
ExpandR: Teaching Dense Retrievers Beyond Queries with LLM Guidance
Paper • 2502.17057 • Published
Interactive Research Guides
Interactive research guides to diffusion decoding, RAG, dense retrieval, long-context DPO, tool use, translation and edge-cloud agents.
- Running
Unmasking Order in Diffusion Language Models
🧩How diffusion language models choose what to unmask next
- Running
RAG Knowledge Conflicts
📚When should a RAG model follow or resist retrieved evidence?
- Running
Query Expansion for Dense Retrieval
📖Compare LLM query expansion and joint retriever training
- Running
Long-Context Preference Optimization
📖Explore bandit chunk sampling for long-context DPO
Selected Research Papers — Pengcheng Huang
Papers on large language models, including diffusion decoding, retrieval-augmented generation, tool use, and edge–cloud agents.
-
Empirical Analysis of Decoding Biases in Masked Diffusion Models
Paper • 2508.13021 • Published -
PIP-KAG: Mitigating Knowledge Conflicts in Knowledge-Augmented Generation via Parametric Pruning
Paper • 2502.15543 • Published • 2 -
SPARK: Adaptive Parametric Knowledge Participation for Faithful Retrieval-Augmented Generation
Paper • 2606.27786 • Published -
ExpandR: Teaching Dense Retrievers Beyond Queries with LLM Guidance
Paper • 2502.17057 • Published
spaces 7
Running
RAG Knowledge Conflicts
📚
When should a RAG model follow or resist retrieved evidence?
Running
LLM Tool Use Generalization
📖
Test how tool trials help agents use unfamiliar tools
Running
Edge-Cloud Agent Routing
📖
Compare routing, delegation, cost and privacy exposure
Running
Terminology-Constrained Translation
📖
Check glossary terms and compare draft and revision
Running
Query Expansion for Dense Retrieval
📖
Compare LLM query expansion and joint retriever training
Running
Long-Context Preference Optimization
📖
Explore bandit chunk sampling for long-context DPO
datasets 8
chengpingan/ParamMute-Training-Data
Viewer • Updated • 32.6k • 32 • 2
chengpingan/ExpandR_retriever_training_data
Viewer • Updated • 709k • 22 • 1
chengpingan/ExpandR_llm_training_data
Viewer • Updated • 30k • 18 • 1
chengpingan/CoFaithfulQA
Viewer • Updated • 8.53k • 307
chengpingan/EC2-Bench
Viewer • Updated • 128 • 2.75k • 1
chengpingan/wildclaw_trej
Updated • 8
chengpingan/LLM-QE-Retriever-Training-Data
Viewer • Updated • 709k • 4
chengpingan/LLM-QE-DPO-Training-Data
Viewer • Updated • 30k • 7