Research
A paper is listed when one of its authors is affiliated with the lab. Data: OpenAlex.
Q-Align: Query-Boundary Multilingual Alignment for LLM Safety GuardrailsAlibaba · Conference on Empirical Methods in Natural Language Processing (EMNLP)Jan 1Doc2Weight: In-Weights Retrieval-Augmented Generation via Parameter InjectionLG AI Research · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP)Jan 1DenseMixer: Improving MoE Post-Training via Precise Router GradientMicrosoft · Conference on Empirical Methods in Natural Language Processing (EMNLP)Jan 1Lost in Mimicry: Verified Training-Free Post-Editing for LLM TranslationeseLG AI Research · Conference on Machine Translation (WMT)Jan 1ChemEval2: A Diagnostic Benchmark for Evaluating Tool-Integrated Reasoning in ChemistryiFlytek · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP)Jan 1RewardPartial: A Completion Potential Evaluation Framework for Truncated Best-of-N ParadigmZhipu AI · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP)Jan 1MemBook: A Hierarchical Reasoning-based Memory Framework for Hallucination-Resistant LLM AgentsAlibaba · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP)Jan 1Quantifying and Mitigating Default Behaviors in Text-to-Image ModelsAlphabet · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP)Jan 1Mem-LM: An LM-style Model for Reinforced Memory SelectionAlphabet · Conference on Empirical Methods in Natural Language Processing (EMNLP)Jan 1Fake It Till You Make It: Training Deep-Search Agents in Synthetic Search EnvironmentsAlibaba · Conference on Empirical Methods in Natural Language Processing (EMNLP)Jan 1MosaicAudio: Multi-Teacher On-Policy Distillation for Controllable Video-to-Audio GenerationAlibaba · Conference on Empirical Methods in Natural Language Processing (EMNLP)Jan 1Orchestrating Hidden-Intermediate Pruning-and-Distill for MoEs SlimmingAlibaba · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP)Jan 1CAPTR: Caption-based Context Pruning for Multimodal Tabular ReasoningCohere · Conference on Empirical Methods in Natural Language Processing (EMNLP)Jan 1When Structure Is Hidden: Measuring and Diagnosing the Compositional Gap in LLMsMeta · Conference on Empirical Methods in Natural Language Processing (EMNLP)Jan 1Phoenix-VL 1.5 Medium: Continued Pre-Training of Sovereign Multimodal AI for On-Premises Public-Sector DeploymentMistral AI · Conference on Empirical Methods in Natural Language Processing (EMNLP)Jan 1Do Benchmarks Measure the Same Thing? Evaluation-Free Benchmark Redundancy Analysis with SAEAlibaba · Conference on Empirical Methods in Natural Language Processing (EMNLP)Jan 1APEX : Adaptively Balancing Off-Policy Guidance and On-Policy Exploration for LLM ReasoningAlibaba · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP)Jan 1SkillLadder: Multi-User Skill Evolution with Compliance-Oriented Synchronization and Quality-Gated SynthesisMicrosoft · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP)Jan 1Silence over Error: Risk-Averse Navigation Landmark Extraction via Progressive MLLM AlignmentAlibaba · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP)Jan 1Unsupervised Domain Specialization for Multimodal EmbeddingsMicrosoft · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP)Jan 1Self-Evolving Vision Language Models for Structured Visual UnderstandingNVIDIA · Conference on Empirical Methods in Natural Language Processing (EMNLP)Jan 1Perception Is the Bottleneck for Large Audio Reasoning ModelsAlibaba · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP)Jan 1When Debate Messages Help or Hurt: Diagnosing Communication in Self-DistillationiFlytek · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP)Jan 1IRD: Steering Reasoning Models with Intent, Risk, and DecisionAlibaba · Conference on Empirical Methods in Natural Language Processing (EMNLP)Jan 1Compressing Only What Is Compressible: Memory-Efficient Multiscale Muon via Selective Low-Rank ProjectionMicrosoft · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP)Jan 1