Rank-1 LoRAs Encode Interpretable Reasoning Signals

• URL: http://arxiv.org/abs/2511.06739v1
• Date: Mon, 10 Nov 2025 06:00:25 GMT
• Title: Rank-1 LoRAs Encode Interpretable Reasoning Signals
• Abstract summary: Reasoning models leverage inference-time compute to significantly enhance the performance of language models on logical tasks.<n>Despite their wide adoption, the mechanisms underpinning the enhanced performance of these reasoning models are not well understood.<n>We show that the majority of new capabilities in reasoning models can be elicited by small, single-rank changes to base model parameters.
• Score: 0.764671395172401
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Reasoning models leverage inference-time compute to significantly enhance the performance of language models on difficult logical tasks, and have become a dominating paradigm in frontier LLMs. Despite their wide adoption, the mechanisms underpinning the enhanced performance of these reasoning models are not well understood. In this work, we show that the majority of new capabilities in reasoning models can be elicited by small, single-rank changes to base model parameters, with many of these changes being interpretable. Specifically, we use a rank-1 LoRA to create a minimal parameter adapter for Qwen-2.5-32B-Instruct which recovers 73-90% of reasoning-benchmark performance compared to a full parameter finetune. We find that the activations of this LoRA are as interpretable as MLP neurons, and fire for reasoning-specific behaviors. Finally, we train a sparse autoencoder on the entire activation state of this LoRA and identify fine-grained and monosemantic features. Our findings highlight that reasoning performance can arise largely from minimal changes to base model parameters, and explore what these changes affect. More broadly, our work shows that parameter-efficient training methods can be used as a targeted lens for uncovering fundamental insights about language model behavior and dynamics.

Related papers

• Weight-Adjusted Gradients Reveal Parameter Importance and Failure Modes in LLMs [12.559788353912651]
  We introduce Weight-Adjusted Gradients (WAG), a simple approach for estimating parameter importance.<n>WAG explicitly captures the interaction between model weights and first-order gradient information.<n>We show that WAG surfaces a tiny but critical subset of parameters whose modification leads to dramatic degradation in performance.
  arXiv  Detail & Related papers  (2026-07-12T15:21:19Z)
• LLM attribution analysis across different fine-tuning strategies and model scales for automated code compliance [1.154460794319478]
  This paper employs a perturbation-based attribution analysis to compare the interpretive behaviors of large language models (LLMs) across different fine-tuning strategies.<n>Our results show that FFT produces attribution patterns that are statistically different and more focused than those from parameter-efficient fine-tuning methods.
  arXiv  Detail & Related papers  (2026-04-16T23:54:26Z)
• High-Rank Structured Modulation for Parameter-Efficient Fine-Tuning [57.85676271833619]
  Low-rank Adaptation (LoRA) uses a low-rank update method to simulate full parameter fine-tuning.<n>We present textbfSMoA, a high-rank textbfStructured textbfMOdulation textbfAdapter that uses fewer trainable parameters while maintaining a higher rank.
  arXiv  Detail & Related papers  (2026-01-12T13:06:17Z)
• Understanding the Effects of Domain Finetuning on LLMs [60.874016669351874]
  We present the first systematic study of domain-specific fine-tuning in large medical language models.<n>Our analysis reveals that fine-tuning modifies only a small subset of the representational subspace.<n>To interpret these changes in subspaces, we propose tuning vectors, which explicitly capture the directional parameter shifts induced by fine-tuning.
  arXiv  Detail & Related papers  (2025-10-10T13:14:06Z)
• AILoRA: Function-Aware Asymmetric Initialization for Low-Rank Adaptation of Large Language Models [11.663809872664105]
  Low-Rank Adaptation (LoRA) has emerged as one of the most widely adopted approaches.<n>LoRA is typically applied to the $WQ$ and $WV$ projection matrices of self-attention modules.<n>We introduce textAILoRA, a novel parameter-efficient method that incorporates function-aware asymmetric low-rank priors.
  arXiv  Detail & Related papers  (2025-10-09T10:13:16Z)
• Stands to Reason: Investigating the Effect of Reasoning on Idiomaticity Detection [2.8330244018167945]
  We examine how reasoning capabilities in Large Language Models affect idiomaticity detection performance.<n>We find the effect of reasoning to be smaller and more varied than expected.<n>For smaller models, producing chain-of-thought (CoT) reasoning increases performance from Math-tuned intermediate models, but not to the levels of the base models.
  arXiv  Detail & Related papers  (2025-08-18T21:17:09Z)
• DiffLoRA: Differential Low-Rank Adapters for Large Language Models [59.58987161199141]
  We introduce DiffLoRA, a parameter-efficient adaptation of the differential attention mechanism, with low-rank adapters on both positive and negative attention terms.<n>We evaluate DiffLoRA across a broad range of NLP tasks, including general benchmarks, many-shot in-context learning, RAG, and long-context tests.
  arXiv  Detail & Related papers  (2025-07-31T14:24:59Z)
• KV Cache Steering for Controlling Frozen LLMs [80.50365534625438]
  cache steering is a lightweight method for implicit steering of language models.<n>We apply cache steering to induce chain-of-thought reasoning in small language models.
  arXiv  Detail & Related papers  (2025-07-11T17:59:36Z)
• Can Reasoning Help Large Language Models Capture Human Annotator Disagreement? [84.32752330104775]
  Variation in human annotation (i.e., disagreements) is common in NLP.<n>We evaluate the influence of different reasoning settings on Large Language Model disagreement modeling.<n>Surprisingly, our results show that RLVR-style reasoning degrades performance in disagreement modeling.
  arXiv  Detail & Related papers  (2025-06-24T09:49:26Z)
• Relative Overfitting and Accept-Reject Framework [5.465098504510676]
  We propose an ensemble framework that governs how models are segmented to ensure performance improvement.<n>We detail the patterns of this framework within the domain of NLP and briefly describe its to other fields, such as computer vision (CV) and AI for science.
  arXiv  Detail & Related papers  (2025-05-12T17:36:14Z)
• Improving Reasoning Performance in Large Language Models via Representation Engineering [2.0099933815960256]
  We propose a representation engineering approach for large language models (LLMs)<n>Model activations are read from the residual stream of an LLM when processing a reasoning task.<n>We show that an LLM can, to a certain degree, be controlled to improve its perceived reasoning ability by modulating activations.
  arXiv  Detail & Related papers  (2025-04-28T04:58:43Z)
• Exploring Model Editing for LLM-based Aspect-Based Sentiment Classification [17.512415475301395]
  We investigate model editing to serve an efficient method for adapting large language models (LLMs) to solve aspect-based sentiment classification.<n>Our findings reveal that a distinct set of mid-layer representations is essential for detecting the sentiment polarity of given aspect words.<n>We develop a model editing approach that focuses exclusively on these critical parts of the LLM, leading to a more efficient method for adapting LLMs.
  arXiv  Detail & Related papers  (2025-03-19T11:21:37Z)
• Learning on LoRAs: GL-Equivariant Processing of Low-Rank Weight Spaces for Large Finetuned Models [38.197552424549514]
  Low-rank adaptations (LoRAs) have revolutionized the finetuning of large foundation models. LoRAs present opportunities for applying machine learning techniques that take these low-rank weights themselves as inputs. In this paper, we investigate the potential of Learning on LoRAs (LoL), a paradigm where LoRA weights serve as input to machine learning models.
  arXiv  Detail & Related papers  (2024-10-05T15:52:47Z)
• MELoRA: Mini-Ensemble Low-Rank Adapters for Parameter-Efficient Fine-Tuning [71.50432879573614]
  Low-rank adaptation (LoRA) is based on the idea that the adaptation process is intrinsically low-dimensional.<n>We present MELoRA, a mini-ensemble low-rank adapters that uses fewer trainable parameters while maintaining a higher rank.<n>Our experimental results show that, compared to LoRA, MELoRA achieves better performance with 8 times fewer trainable parameters on natural language understanding tasks and 36 times fewer trainable parameters on instruction following tasks.
  arXiv  Detail & Related papers  (2024-02-27T07:14:12Z)

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.