Divergence Decoding: Training-Free Capability Fusion

• URL: http://arxiv.org/abs/2607.27248v1
• Date: Tue, 28 Jul 2026 06:52:19 GMT
• Title: Divergence Decoding: Training-Free Capability Fusion
• Abstract summary: Divergence Decoding is a training-free framework for capability fusion.<n>It reconstructs the "draft-and-verify" skeleton of speculative decoding into an adaptive routing mechanism.
• Score: 25.926628015657915
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: While large language models excel in reasoning, these generalists often lack knowledge for specialized scientific domains. Conversely, domain models~(specialists), while knowledgeable, suffer from specialization side-effects including diminished logic and reduced robustness.To address this dilemma, we introduce Divergence Decoding, a training-free framework for capability fusion. It reconstructs the "draft-and-verify" skeleton of speculative decoding into an adaptive routing mechanism. The core is using Jensen-Shannon divergence to monitor the distributional disagreement between the two models at each token. When the specialist exhibits significant divergence, our method identifies it as a potential reasoning risk and instantaneously routes control to the generalist. This allows the dynamic injection of general reasoning while preserving domain expertise, achieving inference-time policy composition of the generalist and the specialist.We evaluate Divergence Decoding across diverse model families (Qwen and Llama series) on challenging scientific benchmarks (GPQA, ChemBench, and ChemCoTBench). Experimental results demonstrate that Divergence Decoding outperforms both the domain-specialized and general-purpose models, effectively surpassing the performance of most single-model baseline. This suggests that Divergence Decoding provides a general, training-free paradigm for fusing diverse LLM capabilities through adaptive inference-time collaboration.

Related papers

• CDFM: Towards a General-Purpose Causal Discovery Foundation Model [62.32138962508216]
  Causal discovery is the process of recovering underlying causal structures from observational data.<n>We formulate the Causal Discovery Foundation Model (CDFM) as a unified, general-purpose framework for zero-shot structural inference.<n>CDFM consistently outperforms traditional algorithms, driving a paradigm shift toward a general-purpose causal discovery foundation model.
  arXiv  Detail & Related papers  (2026-07-13T12:58:28Z)
• Human-Machine Collaboration on Generative Meta-Learning: Model and Algorithm [22.212135025785244]
  Generalizing machine learning models to environments that differ from their training distribution remains a critical hurdle.<n>We propose Generative Meta-Learning with Human Feedback (GMHF), a novel framework that bridges this domain gap by leveraging expert intuition to guide data synthesis.
  arXiv  Detail & Related papers  (2026-07-01T13:29:54Z)
• CausalMoE: A Billion-Scale Multimodal Foundation Model for Granger Causal Discovery with Pattern-Routed Heterogeneous Experts [25.563603697914242]
  We propose CausalMoE, a billion-scale multimodal Granger causal foundation model.<n>CaulMoE introduces a Pattern-Routed Mixture of Heterogeneous Experts, which dynamically identifies latent temporal patterns.<n>We also design a Causality-Aware Self-Attention mechanism operating across variables, yielding sparse Granger causal graphs.
  arXiv  Detail & Related papers  (2026-06-11T07:57:23Z)
• How and Why LLMs Generalize: A Fine-Grained Analysis of LLM Reasoning from Cognitive Behaviors to Low-Level Patterns [51.02752099869218]
  Large Language Models (LLMs) display strikingly different generalization behaviors.<n>We introduce a novel benchmark that decomposes reasoning into atomic core skills.<n>We show that RL-tuned models maintain more stable behavioral profiles and resist collapse in reasoning skills, whereas SFT models exhibit sharper drift and overfit to surface patterns.
  arXiv  Detail & Related papers  (2025-12-30T08:16:20Z)
• Nemotron-Cascade: Scaling Cascaded Reinforcement Learning for General-Purpose Reasoning Models [71.9060068259379]
  We propose cascaded domain-wise reinforcement learning to build general-purpose reasoning models.<n>Our 14B model, after RL, outperforms its SFT teacher, DeepSeek-R1-0528, on LiveCodeBench v5/v6 Pro and silver-medal performance in the 2025 International Olympiad in Informatics (IOI)
  arXiv  Detail & Related papers  (2025-12-15T18:02:35Z)
• Deep Equilibrium models for Poisson Imaging Inverse problems via Mirror Descent [7.248102801711294]
  Deep Equilibrium Models (DEQs) are implicit neural networks with fixed points.<n>We introduce a novel DEQ formulation based on Mirror Descent defined in terms of a tailored non-Euclidean geometry.<n>We propose computational strategies that enable both efficient training and fully parameter-free inference.
  arXiv  Detail & Related papers  (2025-07-15T16:33:01Z)
• Identifying Causal Direction via Variational Bayesian Compression [6.928582707713723]
  A key principle is the algorithmic Markov condition, which postulates that the joint distribution, when factorized according to the causal direction, yields a more succinct codelength.<n>Previous approaches approximate these codelengths by relying on simple functions or Gaussian processes (GPs) with easily evaluable complexity.<n>We propose leveraging the variational Bayesian learning of neural networks as an interpretation of the codelengths.
  arXiv  Detail & Related papers  (2025-05-12T12:40:15Z)
• More Experts Than Galaxies: Conditionally-overlapping Experts With Biologically-Inspired Fixed Routing [5.846028298833611]
  Conditionally Overlapping Mixture of ExperTs (COMET) is a general deep learning method that inducing a modular, sparse architecture with an exponential number of overlapping experts.<n>We demonstrate the effectiveness of COMET on a range of tasks, including image classification, language modeling, and regression.
  arXiv  Detail & Related papers  (2024-10-10T14:58:18Z)
• LoRKD: Low-Rank Knowledge Decomposition for Medical Foundation Models [59.961172635689664]
  "Knowledge Decomposition" aims to improve the performance on specific medical tasks. We propose a novel framework named Low-Rank Knowledge Decomposition (LoRKD) LoRKD explicitly separates gradients from different tasks by incorporating low-rank expert modules and efficient knowledge separation convolution.
  arXiv  Detail & Related papers  (2024-09-29T03:56:21Z)
• ER: Equivariance Regularizer for Knowledge Graph Completion [107.51609402963072]
  We propose a new regularizer, namely, Equivariance Regularizer (ER) ER can enhance the generalization ability of the model by employing the semantic equivariance between the head and tail entities. The experimental results indicate a clear and substantial improvement over the state-of-the-art relation prediction methods.
  arXiv  Detail & Related papers  (2022-06-24T08:18:05Z)
• On the Generalization and Adaption Performance of Causal Models [99.64022680811281]
  Differentiable causal discovery has proposed to factorize the data generating process into a set of modules. We study the generalization and adaption performance of such modular neural causal models. Our analysis shows that the modular neural causal models outperform other models on both zero and few-shot adaptation in low data regimes.
  arXiv  Detail & Related papers  (2022-06-09T17:12:32Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.