Opening the Black Box: Interpretable LLMs via Semantic Resonance Architecture

• URL: http://arxiv.org/abs/2509.14255v1
• Date: Fri, 12 Sep 2025 09:02:48 GMT
• Title: Opening the Black Box: Interpretable LLMs via Semantic Resonance Architecture
• Abstract summary: Mixture-of-Experts (MoE) models improve efficiency through sparse activation, yet typically rely on opaque, learned gating functions.<n>We introduce the Semantic Resonance Architecture (SRA), an MoE approach designed to ensure that routing decisions are inherently interpretable.<n>SRA replaces learned gating with a Chamber of Semantic Resonance (CSR) module, which routes tokens based on cosine similarity with trainable semantic anchors.
• Score: 0.0
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Large language models (LLMs) achieve remarkable performance but remain difficult to interpret. Mixture-of-Experts (MoE) models improve efficiency through sparse activation, yet typically rely on opaque, learned gating functions. While similarity-based routing (Cosine Routers) has been explored for training stabilization, its potential for inherent interpretability remains largely untapped. We introduce the Semantic Resonance Architecture (SRA), an MoE approach designed to ensure that routing decisions are inherently interpretable. SRA replaces learned gating with a Chamber of Semantic Resonance (CSR) module, which routes tokens based on cosine similarity with trainable semantic anchors. We also introduce a novel Dispersion Loss that encourages orthogonality among anchors to enforce diverse specialization. Experiments on WikiText-103 demonstrate that SRA achieves a validation perplexity of 13.41, outperforming both a dense baseline (14.13) and a Standard MoE baseline (13.53) under matched active parameter constraints (29.0M). Crucially, SRA exhibits superior expert utilization (1.0% dead experts vs. 14.8% in the Standard MoE) and develops distinct, semantically coherent specialization patterns, unlike the noisy specialization observed in standard MoEs. This work establishes semantic routing as a robust methodology for building more transparent and controllable language models.

Related papers

• DOT-MoE: Differentiable Optimal Transport for MoEfication [11.183649354845885]
  DOT-MoE is a novel framework that formulates the decomposition of dense layers as a Differentiable Optimal Transport problem.<n>We demonstrate that DOT-MoE significantly outperforms structured pruning, clustering, and random-split baselines.
  arXiv  Detail & Related papers  (2026-06-01T04:19:16Z)
• Self-Distilled Trajectory-Aware Boltzmann Modeling: Bridging the Training-Inference Discrepancy in Diffusion Language Models [65.89572755202245]
  Diffusion Language Models (DLMs) offer stronger global awareness and highly parallel generation.<n>Post-training DLMs with standard Negative Evidence Lower Bound (NELBO)-based supervised fine-tuning remains inefficient.<n>We propose a self-distilled trajectory-based post-training framework that aligns training with the easy-to-hard structure of inference.
  arXiv  Detail & Related papers  (2026-05-12T09:39:06Z)
• Too Correct to Learn: Reinforcement Learning on Saturated Reasoning Data [55.84428098924793]
  We propose Constrained Uniform Top-K Sampling (CUTS), a parameter-free decoding strategy enforcing structure-preserving exploration.<n>We integrate this into Mixed-CUTS, a training framework synergizing exploitative and exploratory rollouts to amplify intra-group advantage variance.<n> Notably, Mixed-CUTS improves Pass@1 accuracy on the challenging AIME25 benchmark by up to 15.1% over standard GRPO.
  arXiv  Detail & Related papers  (2026-04-20T16:43:28Z)
• LaSER: Internalizing Explicit Reasoning into Latent Space for Dense Retrieval [74.72139580745511]
  LaSER is a novel self-distillation framework that internalizes explicit reasoning into the latent space of retrievers.<n>Our method successfully combines the reasoning depth of explicit CoT pipelines with the inference efficiency of standard dense retrievers.
  arXiv  Detail & Related papers  (2026-03-02T04:11:18Z)
• Latent-Space Contrastive Reinforcement Learning for Stable and Efficient LLM Reasoning [16.244366307890832]
  We propose textbfDeepLatent Reasoning (DLR), a latent-space bidirectional contrastive reinforcement learning framework.<n>This framework shifts the trial-and-error cost from expensive token-level full sequence generation to the continuous latent manifold.<n> Experiments demonstrate that DLR achieves more stable training convergence, supports longer-horizon reasoning chains, and facilitates the sustainable accumulation of reasoning capabilities.
  arXiv  Detail & Related papers  (2026-01-24T03:18:22Z)
• Routing Matters in MoE: Scaling Diffusion Transformers with Explicit Routing Guidance [79.21541758879012]
  Mixture-of-Experts (MoE) has emerged as a powerful paradigm for scaling model capacity while preserving computational efficiency.<n>We present ProMoE, an MoE framework featuring a two-step router with explicit routing guidance that promotes expert specialization.
  arXiv  Detail & Related papers  (2025-10-28T17:59:02Z)
• Joint Learning using Mixture-of-Expert-Based Representation for Enhanced Speech Generation and Robust Emotion Recognition [54.44798086835314]
  Speech emotion recognition (SER) plays a critical role in building emotion-aware speech systems, but its performance degrades significantly under noisy conditions.<n>We propose the Sparse Mixture-of-Experts Representation Integration Technique (Sparse MERIT), a flexible MTL framework that applies frame-wise expert routing over self-supervised speech representations.<n> Experiments on the MSP-Podcast corpus show that Sparse MERIT consistently outperforms baseline models on both SER and SE tasks.
  arXiv  Detail & Related papers  (2025-09-10T10:18:56Z)
• MoSEs: Uncertainty-Aware AI-Generated Text Detection via Mixture of Stylistics Experts with Conditional Thresholds [50.94175872742232]
  We propose the Mixture of Stylistic Experts (MoSEs) framework that enables stylistics-aware quantification uncertainty.<n>MoSEs contains three core components, namely, the Stylistics Reference Repository (SRR), the Stylistics-Aware Router (SAR), and the Conditional Threshold Estimator (CTE)<n>With a discrimination score, MoSEs yields prediction labels with the corresponding confidence level.
  arXiv  Detail & Related papers  (2025-09-02T16:51:43Z)
• Mixture of Experts Made Intrinsically Interpretable [34.36996159677674]
  We present textbfMoE-X, a Mixture-of-Experts (MoE) language model designed to be emphintrinsically interpretable.<n>Our approach is motivated by the observation that, in language models, wider networks with sparse activations are more likely to capture interpretable factors.<n>MoE-X achieves perplexity better than GPT-2, with interpretability surpassing even sparse autoencoder (SAE)-based approaches.
  arXiv  Detail & Related papers  (2025-03-05T17:40:54Z)
• Unchosen Experts Can Contribute Too: Unleashing MoE Models' Power by Self-Contrast [58.98411447739218]
  Mixture-of-Experts (MoE) has emerged as a prominent architecture for scaling model size while maintaining computational efficiency. We propose Self-Contrast Mixture-of-Experts (SCMoE), a training-free strategy that utilizes unchosen experts in a self-contrast manner during inference. Our method is conceptually simple and computationally lightweight, as it incurs minimal latency compared to greedy decoding.
  arXiv  Detail & Related papers  (2024-05-23T12:45:29Z)
• Mixture of insighTful Experts (MoTE): The Synergy of Thought Chains and Expert Mixtures in Self-Alignment [103.05005690990271]
  Mixture of insighTful Experts (MoTE) is a novel framework that combines reasoning chains and expert mixtures to improve self-alignments.<n>MoTE significantly improves model safety, jailbreak resistance, and over-refusal capabilities, achieving performance comparable to OpenAI's state-of-the-art o1 model.
  arXiv  Detail & Related papers  (2024-05-01T15:06:05Z)
• Soft Merging of Experts with Adaptive Routing [38.962451264172856]
  We introduce Soft Merging of Experts with Adaptive Routing (SMEAR) SMEAR avoids discrete routing by using a single "merged" expert constructed via a weighted average of all of the experts' parameters. We empirically validate that models using SMEAR outperform models that route based on metadata or learn sparse routing through gradient estimation.
  arXiv  Detail & Related papers  (2023-06-06T15:04:31Z)

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.