A bipartite factor graph with message-passing protocol and asymmetric damping aggregates multi-LLM predictions, cutting token use by 97% and API calls by 6X while outperforming baselines on MMLU, MMLU-Pro, GPQA, and MedMCQA.
hub Canonical reference
arXiv preprint arXiv:2403.03883 , year=
Canonical reference. 100% of citing Pith papers cite this work as background.
16
Pith papers citing it
18
external citations · external index
Background
100% of classified citations
hub tools
citation-role summary
background 5
citation-polarity summary
roles
background 5polarities
background 5representative citing papers
VertMark embeds robust, training-free watermarks into vertical domain language models by creating hidden semantic equivalence between low-frequency triggers and high-frequency domain terms via parameter swaps, supporting reliable verification with negligible performance impact.
DECO is a sparse MoE architecture with ReLU-based routing, learnable expert scaling, and NormSiLU activation that matches dense Transformer performance at 20% expert activation and delivers 2.93x speedup on Jetson AGX Orin.
EvoRAG adds a feedback-driven backpropagation step that attributes response quality to individual knowledge-graph triplets and updates the graph to raise reasoning accuracy by 7.34 percent over prior KG-RAG methods.
OpsLLM's pipeline (HITL data curation, SFT, GRPO RL with a domain process reward model) improves LLM accuracy on software-operations QA and RCA, especially on in-distribution root-cause-analysis tasks.
CAMEL is a scaling law capturing nonlinear model-size and mixture interactions to extrapolate optimal data mixtures for large LLMs from small-model experiments, reducing optimization cost by 50% and improving benchmarks by up to 3%.
The survey organizes RAG methods via a taxonomy of query-based, logits-based, latent, and parametric fusion with comparisons on accessibility, efficiency, applications, and challenges.
Reasoning-oriented LLMs reach up to 0.91 quadratic weighted kappa agreement with experts on public law cases when given sample solutions and grading rubrics, but only 0.60 on criminal law cases.
Domain-trained small language model Olava Extract outperforms frontier LLMs on structured contract extraction with macro F1 0.812, micro F1 0.842, highest precision, and 78-97% lower inference cost.
ChipLingo trains LLMs on EDA data via corpus construction, domain-adaptive pretraining, and RAG scenario alignment, reaching 59.7% accuracy with an 8B model and 70.02% with a 32B model on a new internal EDA benchmark.
Maat is a ReAct agent that orchestrates tools and RAG for competition law research, outperforming baselines on case-specific tasks while providing official citations.
The paper introduces a taxonomy of AI safety for LLMs organized into Trustworthy AI, Responsible AI, and Safe AI perspectives, accompanied by a review of state-of-the-art methods, challenges, and future directions.
Explicit memory modeled on the hippocampus is the cornerstone needed to advance LLMs to AGI because their implicit statistical learning cannot produce higher cognitive functions.
Proposes Knowledge Objects to externalize implicit AI knowledge for human verification, addressing a gap in current reliability methods.
LLM pipeline for joint post-level assessment and user-level temporal modeling of mental health from ordered social media posts in a shared task.
citing papers explorer
-
From Talking Words to Sharing Thoughts: Scalable Multi-LLM Aggregation via Structured Message Passing
A bipartite factor graph with message-passing protocol and asymmetric damping aggregates multi-LLM predictions, cutting token use by 97% and API calls by 6X while outperforming baselines on MMLU, MMLU-Pro, GPQA, and MedMCQA.
-
VertMark: A Unified Training-Free Robust Watermarking Framework for Vertical Domain Pre-trained Language Models
VertMark embeds robust, training-free watermarks into vertical domain language models by creating hidden semantic equivalence between low-frequency triggers and high-frequency domain terms via parameter swaps, supporting reliable verification with negligible performance impact.
-
DECO: Sparse Mixture-of-Experts with Dense-Comparable Performance on End-Side Devices
DECO is a sparse MoE architecture with ReLU-based routing, learnable expert scaling, and NormSiLU activation that matches dense Transformer performance at 20% expert activation and delivers 2.93x speedup on Jetson AGX Orin.
-
EvoRAG: Making Knowledge Graph-based RAG Automatically Evolve through Feedback-driven Backpropagation
EvoRAG adds a feedback-driven backpropagation step that attributes response quality to individual knowledge-graph triplets and updates the graph to raise reasoning accuracy by 7.34 percent over prior KG-RAG methods.
-
OpsLLM: Construction of Large Language Model for Software Operations with Multi-stage Learning
OpsLLM's pipeline (HITL data curation, SFT, GRPO RL with a domain process reward model) improves LLM accuracy on software-operations QA and RCA, especially on in-distribution root-cause-analysis tasks.
-
Capacity-Aware Mixture Law Enables Efficient LLM Data Optimization
CAMEL is a scaling law capturing nonlinear model-size and mixture interactions to extrapolate optimal data mixtures for large LLMs from small-model experiments, reducing optimization cost by 50% and improving benchmarks by up to 3%.
-
Retrieval-Augmented Generation for Natural Language Processing: A Survey
The survey organizes RAG methods via a taxonomy of query-based, logits-based, latent, and parametric fusion with comparisons on accessibility, efficiency, applications, and challenges.
-
GradeLegal: Automated Grading for German Legal Cases
Reasoning-oriented LLMs reach up to 0.91 quadratic weighted kappa agreement with experts on public law cases when given sample solutions and grading rubrics, but only 0.60 on criminal law cases.
-
A Few Good Clauses: Comparing LLMs vs Domain-Trained Small Language Models on Structured Contract Extraction
Domain-trained small language model Olava Extract outperforms frontier LLMs on structured contract extraction with macro F1 0.812, micro F1 0.842, highest precision, and 78-97% lower inference cost.
-
ChipLingo: A Systematic Training Framework for Large Language Models in EDA
ChipLingo trains LLMs on EDA data via corpus construction, domain-adaptive pretraining, and RAG scenario alignment, reaching 59.7% accuracy with an 8B model and 70.02% with a 32B model on a new internal EDA benchmark.
-
Maat: The Agentic Legal Research Assistant for Competition Protection
Maat is a ReAct agent that orchestrates tools and RAG for competition law research, outperforming baselines on case-specific tasks while providing official citations.
-
AI Safety Landscape for Large Language Models: Taxonomy, State-of-the-art, and Future Directions
The paper introduces a taxonomy of AI safety for LLMs organized into Trustworthy AI, Responsible AI, and Safe AI perspectives, accompanied by a review of state-of-the-art methods, challenges, and future directions.
-
Position: Hippocampal Explicit Memory Is the Cornerstone for AGI
Explicit memory modeled on the hippocampus is the cornerstone needed to advance LLMs to AGI because their implicit statistical learning cannot produce higher cognitive functions.
-
Reliable AI Needs to Externalize Implicit Knowledge: A Human-AI Collaboration Perspective
Proposes Knowledge Objects to externalize implicit AI knowledge for human verification, addressing a gap in current reliability methods.
-
Team MKC at CLPsych 2026: Capturing and Characterizing Mental Health Changes through Social Media Timeline Dynamics
LLM pipeline for joint post-level assessment and user-level temporal modeling of mental health from ordered social media posts in a shared task.
- Citation Grounding Measures the Oracle: Graph Coverage Determines Reported LLM Hallucination Rates in Law