{"slug": "dllg-dynamic-logit-level-gating-of-llm-experts", "title": "DLLG: Dynamic Logit-Level Gating of LLM Experts", "summary": "Researchers have introduced DLLG (Dynamic Logit-Level Gating), a framework that learns token-level fusion weights for combining multiple specialized large language models without requiring token-level labels or expert retraining. The method uses a lightweight gating module to predict step-wise fusion weights from sparse response-level supervision, outperforming routing, heuristic ensembling, and parameter-merging baselines across reasoning and code benchmarks. This approach establishes learned logit-level fusion as a scalable paradigm for integrating specialized LLMs while avoiding the trade-offs between adaptability and stability seen in existing methods.", "body_md": "arXiv:2606.04378v1 Announce Type: new\nAbstract: Leveraging multiple specialized LLMs can combine complementary strengths, but existing approaches trade adaptability for stability: routing commits prematurely, heuristic ensembling depends on fragile proxies, and parameter merging introduces interference. We propose DLLG (Dynamic Logit-Level Gating), a dynamic logit-level ensembling framework that learns token-level expert fusion from sparse response-level supervision. A lightweight gating module predicts step-wise fusion weights, linking trajectory-level correctness to generation without token-level labels or expert retraining. Across diverse reasoning and code benchmarks, DLLG consistently outperforms strong routing, heuristic ensembling, and parameter-merging baselines across model scales, highlighting learned logit-level fusion as a robust and scalable paradigm for integrating specialized experts.", "url": "https://wpnews.pro/news/dllg-dynamic-logit-level-gating-of-llm-experts", "canonical_source": "https://arxiv.org/abs/2606.04378", "published_at": "2026-06-04 04:00:00+00:00", "updated_at": "2026-06-04 04:22:35.987012+00:00", "lang": "en", "topics": ["large-language-models", "machine-learning", "artificial-intelligence", "ai-research", "neural-networks"], "entities": ["DLLG"], "alternates": {"html": "https://wpnews.pro/news/dllg-dynamic-logit-level-gating-of-llm-experts", "markdown": "https://wpnews.pro/news/dllg-dynamic-logit-level-gating-of-llm-experts.md", "text": "https://wpnews.pro/news/dllg-dynamic-logit-level-gating-of-llm-experts.txt", "jsonld": "https://wpnews.pro/news/dllg-dynamic-logit-level-gating-of-llm-experts.jsonld"}}