LLMs
Latest breakthroughs in Large Language Models
Articles
R2VC: Modular Fact-Checking with Retrieval, Verification, and⭐8
R2VC is a modular fact-checking pipeline combining retrieval, verification, and confidence calibration, achieving 13.74% higher accuracy on FEVER with reliable ...
Occamy-1.0: Open Pareto-Frontier 35B Intelligence for Co-work⭐8
Occamy-1.0 is a fully open 35B-parameter language model on the Pareto frontier, optimized for collaborative AI co-work with strong reasoning and efficient deplo...
Automating QUBO Formulation Generation from Natural Language⭐8
A multi-agent framework automatically converts natural-language problem descriptions into QUBO formulations, achieving 68% accuracy on the new QUBOBench benchma...
A Multi-Stage Rule-Chaining Framework for Compositional and⭐7
A multi-stage rule-chaining framework for ARC-AGI-2 achieving over 95% accuracy via interpretable symbolic, compositional, and hierarchical cognitive reasoning.
CMNIE: An Information Extraction Benchmark for Chinese Military News⭐9
CMNIE introduces a Chinese military news information extraction benchmark with 13,000 annotated instances, evaluating supervised models and LLMs on joint event,...
NCP-ArchPreview: A Latent-Space Language Model via Next Concept⭐8
NCP-ArchPreview introduces Next Concept Prediction, a latent-space language model that moves beyond next-token prediction toward concept-level pretraining.
Subagents vs. Agent Skills: How to Execute Reusable Knowledge⭐8
Subagent execution outperforms loading agent skills into context for long-horizon tasks. New arXiv research shows how organizing reusable knowledge as subagents...
Adaptive Entangled Game Modules: A Probability-Wave Path Toward⭐8
A probability-wave framework for adaptive agents explains 82–94% of trading decisions, offering a path toward more compact, human-like AGI architectures.
Data-Efficient Language Modeling: From Frontier Advancement to⭐7
Qiushi Engine's three-stage autonomous research on BabyLM 2026 Strict-Small discovers data-efficient learning principles and improves model performance via prin...
X-CoSD: Communication-Efficient Cross-Vocabulary Collaborative⭐8
X-CoSD enables lossless, communication-efficient collaborative speculative decoding across heterogeneous SLM–LLM vocabularies, improving token generation speed ...