ILRR: Inference-Time Steering Method for Masked Diffusion Language Models

• URL: http://arxiv.org/abs/2601.21647v1
• Date: Thu, 29 Jan 2026 12:48:59 GMT
• Title: ILRR: Inference-Time Steering Method for Masked Diffusion Language Models
• Abstract summary: We introduce Iterative Latent Representation Refinement (ILRR), a learning-free framework for steering DLMs using a single reference sequence.<n>ILRR guides generation by dynamically aligning the internal activations of the generated sequence with those of a given reference throughout the denoising process.<n>We further introduce Spatially Modulated Steering, an extension that enables steering long texts using shorter references by regulating guidance intensity across the sequence.
• Score: 8.458339111154585
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Discrete Diffusion Language Models (DLMs) offer a promising non-autoregressive alternative for text generation, yet effective mechanisms for inference-time control remain relatively underexplored. Existing approaches include sampling-level guidance procedures or trajectory optimization mechanisms. In this work, we introduce Iterative Latent Representation Refinement (ILRR), a learning-free framework for steering DLMs using a single reference sequence. ILRR guides generation by dynamically aligning the internal activations of the generated sequence with those of a given reference throughout the denoising process. This approach captures and transfers high-level semantic properties, with a tunable steering scale enabling flexible control over attributes such as sentiment. We further introduce Spatially Modulated Steering, an extension that enables steering long texts using shorter references by regulating guidance intensity across the sequence. Empirically, we demonstrate that ILRR achieves effective attribute steering on LLaDA and MDLM architectures with a minor computational overhead, requiring only one additional parallel forward pass per denoising step. Under the same compute budget, ILRR improves attribute accuracy over comparable baselines by 10$\%$ to 60$\%$ points, while maintaining high generation quality.

Related papers

• DLM-SWAI: Steering Diffusion Language Models Before They Unmask [5.276802062716119]
  DLM-SWAI is a training-free steering method that biases the token distribution at each denoising step using pre-computed token-level style scores.<n> Experiments on style and safety control tasks show that DLM-SWAI effectively steers diffusion language models while preserving generation quality and requiring minimal computational overhead.
  arXiv  Detail & Related papers  (2026-05-28T09:00:14Z)
• AHPA: Adaptive Hierarchical Prior Alignment for Diffusion Transformers [24.937985157569823]
  We argue that such timestep-agnostic alignment is suboptimal because the useful granularity of representation supervision changes systematically with the signal-to-noise ratio.<n>In high-noise regimes, diffusion models benefit more from coarse semantic and layout-level anchoring, whereas in low-noise regimes, the training signal should emphasize spatially detailed and structurally faithful refinement.<n>This non-stationary alignment behavior creates a representational mismatch for static single-level supervisors.
  arXiv  Detail & Related papers  (2026-05-05T03:07:29Z)
• Local Linearity of LLMs Enables Activation Steering via Model-Based Linear Optimal Control [2.752817022620644]
  Inference-time LLM alignment methods offer an alternative to fine-tuning by directly modifying activations during generation.<n>We show that layer-wise dynamics across multiple LLM architectures and scales are well-approximated by locally-linear models.<n>We derive theoretical bounds on setpoint tracking error, enabling formal guarantees on steering performance.
  arXiv  Detail & Related papers  (2026-04-21T03:09:46Z)
• Adaptive Guidance for Retrieval-Augmented Masked Diffusion Models [60.25003813232697]
  We propose Adaptive Retrieval-Augmented Masked Diffusion (ARAM) as a training-free adaptive guidance framework for Masked Diffusion Models (MDMs)<n>ARAM calibrates the guidance scale during denoising according to the Signal-to-Noise Ratio (SNR) of the distributional shift induced by retrieved context.<n>Experiments on multiple knowledge-intensive QA benchmarks show that ARAM improves overall QA performance over competitive RAG baselines.
  arXiv  Detail & Related papers  (2026-03-18T12:54:50Z)
• DLLM Agent: See Farther, Run Faster [94.74432470237817]
  Diffusion large language models (DLLMs) have emerged as an alternative to autoregressive (AR) decoding with appealing efficiency and modeling properties.<n>We study this in a controlled setting by instantiatingDLLM and AR backbones within the same agent workflow.<n>We find thatDLLM Agents are on average over 30% faster end to end than AR agents, with some cases exceeding 8x speedup.
  arXiv  Detail & Related papers  (2026-02-07T09:01:18Z)
• Merging Beyond: Streaming LLM Updates via Activation-Guided Rotations [55.047454145941366]
  Streaming Merging is an innovative model updating paradigm that conceptualizes merging as an iterative optimization process.<n> ARM is a strategy designed to approximate gradient descent dynamics.<n> ARM requires only early SFT checkpoints and, through iterative merging, surpasses the fully converged SFT model.
  arXiv  Detail & Related papers  (2026-02-03T08:15:57Z)
• RISER: Orchestrating Latent Reasoning Skills for Adaptive Activation Steering [62.63376387138257]
  We propose a plug-and-play intervention framework that adaptively steers large language models (LLMs) reasoning in activation space.<n>RISER constructs a library of reusable reasoning vectors and employs a lightweight Router to dynamically compose them for each input.<n>The Router is optimized via reinforcement learning under task-level rewards, activating latent cognitive primitives in an emergent and compositional manner.
  arXiv  Detail & Related papers  (2026-01-14T08:04:33Z)
• Activation Steering for Masked Diffusion Language Models [1.0980666029958932]
  Masked diffusion language models generate text through an iterative denoising process.<n>We present an activation-steering framework for MDLMs that computes layer-wise steering vectors from a single forward pass.<n>Experiments on LLaDA-8B-Instruct demonstrate reliable modulation of high-level attributes.
  arXiv  Detail & Related papers  (2025-12-30T11:10:52Z)
• PEFA-AI: Advancing Open-source LLMs for RTL generation using Progressive Error Feedback Agentic-AI [5.455262834289454]
  We present an agentic flow consisting of multiple agents that collaboratively complete the task of Register Transfer Level (RTL) generation without human intervention.<n>A key feature of the proposed flow is the progressive error feedback system of agents (PEFA), a self-correcting mechanism.<n>To validate this adaptive approach to code generation, benchmarking is performed using two opensource natural language-to-RTL datasets.
  arXiv  Detail & Related papers  (2025-11-06T00:19:47Z)
• In-Distribution Steering: Balancing Control and Coherence in Language Model Generation [0.0815557531820863]
  We introduce In-Distribution Steering (IDS), a novel method that adapts steering strength based on the input data distribution in representation space.<n>IDS achieves strong accuracy on classification tasks while producing coherent text without collapse, making IDS particularly well suited for real-world applications.
  arXiv  Detail & Related papers  (2025-10-15T08:31:37Z)
• Beyond Next-Token Prediction: A Performance Characterization of Diffusion versus Autoregressive Language Models [82.87985794856803]
  Large Language Models (LLMs) have achieved state-of-the-art performance on a broad range of Natural Language Processing (NLP) tasks.<n>Recently, Diffusion Language Models (DLMs) have emerged as a promising alternative architecture.
  arXiv  Detail & Related papers  (2025-10-05T10:50:52Z)
• Multimodal LLM-Guided Semantic Correction in Text-to-Image Diffusion [52.315729095824906]
  MLLM Semantic-Corrected Ping-Pong-Ahead Diffusion (PPAD) is a novel framework that introduces a Multimodal Large Language Model (MLLM) as a semantic observer during inference.<n>It performs real-time analysis on intermediate generations, identifies latent semantic inconsistencies, and translates feedback into controllable signals that actively guide the remaining denoising steps.<n>Extensive experiments demonstrate PPAD's significant improvements.
  arXiv  Detail & Related papers  (2025-05-26T14:42:35Z)
• LLM Braces: Straightening Out LLM Predictions with Relevant Sub-Updates [27.022532404557264]
  We propose LLMBRACES, a method that computes relevance scores associated with value vectors in FFN layers.<n>By optimizing sub-update contributions, LLMBRACES refines the prediction process, leading to more accurate and reliable outputs.<n>LLMBRACES excels in sentiment-controlled generation and toxicity reduction, highlighting its potential for flexible, controlled text generation across applications.
  arXiv  Detail & Related papers  (2025-03-20T16:55:26Z)
• Refining Sentence Embedding Model through Ranking Sentences Generation with Large Language Models [60.00178316095646]
  Sentence embedding is essential for many NLP tasks, with contrastive learning methods achieving strong performance using datasets like NLI.<n>Recent studies leverage large language models (LLMs) to generate sentence pairs, reducing annotation dependency.<n>We propose a method for controlling the generation direction of LLMs in the latent space. Unlike unconstrained generation, the controlled approach ensures meaningful semantic divergence.<n> Experiments on multiple benchmarks demonstrate that our method achieves new SOTA performance with a modest cost in ranking sentence synthesis.
  arXiv  Detail & Related papers  (2025-02-19T12:07:53Z)
• LF-Steering: Latent Feature Activation Steering for Enhancing Semantic Consistency in Large Language Models [16.37602070339033]
  Large Language Models (LLMs) often generate inconsistent responses when prompted with semantically equivalent paraphrased inputs.<n>We propose LF-Steering, a novel activation steering approach to precisely identify latent feature representations responsible for semantic inconsistency.<n>Our method maps the hidden states of the relevant transformer layer into a sparsely activated, high-dimensional feature space based on a sparse autoencoder.
  arXiv  Detail & Related papers  (2025-01-19T13:06:51Z)

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.