TABES: Trajectory-Aware Backward-on-Entropy Steering for Masked Diffusion Models

• URL: http://arxiv.org/abs/2602.00250v1
• Date: Fri, 30 Jan 2026 19:10:32 GMT
• Title: TABES: Trajectory-Aware Backward-on-Entropy Steering for Masked Diffusion Models
• Authors: Shreshth Saini, Avinab Saha, Balu Adsumilli, Neil Birkbeck, Yilin Wang, Alan C. Bovik,
• Abstract summary: Backward-on-Entropy (BoE) Steering is a gradient-guided inference framework that approximates infinite-horizon context via a single backward pass.<n>To ensure scalability, we introduce ttexttActiveQueryAttention, a sparse adjoint primitive that exploits the structure of the masking objective to reduce backward pass complexity.
• Score: 35.327100592206115
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Masked Diffusion Models (MDMs) have emerged as a promising non-autoregressive paradigm for generative tasks, offering parallel decoding and bidirectional context utilization. However, current sampling methods rely on simple confidence-based heuristics that ignore the long-term impact of local decisions, leading to trajectory lock-in where early hallucinations cascade into global incoherence. While search-based methods mitigate this, they incur prohibitive computational costs ($O(K)$ forward passes per step). In this work, we propose Backward-on-Entropy (BoE) Steering, a gradient-guided inference framework that approximates infinite-horizon lookahead via a single backward pass. We formally derive the Token Influence Score (TIS) from a first-order expansion of the trajectory cost functional, proving that the gradient of future entropy with respect to input embeddings serves as an optimal control signal for minimizing uncertainty. To ensure scalability, we introduce \texttt{ActiveQueryAttention}, a sparse adjoint primitive that exploits the structure of the masking objective to reduce backward pass complexity. BoE achieves a superior Pareto frontier for inference-time scaling compared to existing unmasking methods, demonstrating that gradient-guided steering offers a mathematically principled and efficient path to robust non-autoregressive generation. We will release the code.

Related papers

• Sample- and Hardware-Efficient Fidelity Estimation by Stripping Phase-Dominated Magic [0.2796197251957245]
  This work proposes a sample- and gate-efficient fidelity estimation algorithm that is affordable within feasible quantum devices.<n>We show that the fidelity estimation with pure states close to the structure of phase states, for which sample-efficient DFE is limited by their strong entanglement and magic, can be done by using $mathcalO(mathrmpoly(n))$ sampling copies, with a single $n$-qubit fan-out gate.
  arXiv  Detail & Related papers  (2026-02-10T12:13:25Z)
• Euphonium: Steering Video Flow Matching via Process Reward Gradient Guided Stochastic Dynamics [49.242224984144904]
  We propose Euphonium, a novel framework that steers generation via process reward gradient guided dynamics.<n>Our key insight is to formulate the sampling process as a theoretically principled algorithm that explicitly incorporates the gradient of a Process Reward Model.<n>We derive a distillation objective that internalizes the guidance signal into the flow network, eliminating inference-time dependency on the reward model.
  arXiv  Detail & Related papers  (2026-02-04T08:59:57Z)
• 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)
• PROMISE: Process Reward Models Unlock Test-Time Scaling Laws in Generative Recommendations [52.67948063133533]
  Generative Recommendation has emerged as a promising paradigm, reformulating recommendation as a sequence-to-sequence generation task over hierarchical Semantic IDs.<n>Existing methods suffer from a critical issue we term Semantic Drift, where errors in early, high-level tokens irreversibly divert the generation trajectory into irrelevant semantic subspaces.<n>We propose Promise, a novel framework that integrates dense, step-by-step verification into generative models.
  arXiv  Detail & Related papers  (2026-01-08T07:38:46Z)
• Revisiting Weighted Strategy for Non-stationary Parametric Bandits and MDPs [56.246783503873225]
  This paper revisits the weighted strategy for non-stationary parametric bandits.<n>We propose a simpler weight-based algorithm that is as efficient as window/restart-based algorithms.<n>Our framework can be used to improve regret bounds of other parametric bandits.
  arXiv  Detail & Related papers  (2026-01-03T04:50:21Z)
• Learning Unmasking Policies for Diffusion Language Models [33.44995119635116]
  Language Models (dLLMs) now match the downstream performance of their autoregressive counterparts on many tasks.<n>One particularly successful variant is masked discrete diffusion, in which a buffer filled with special mask tokens is progressively replaced with tokens sampled from the model's vocabulary.<n>In this work, we propose to train sampling procedures using reinforcement learning.
  arXiv  Detail & Related papers  (2025-12-09T20:44:33Z)
• On Geometric Structures for Policy Parameterization in Continuous Control [7.056222499095849]
  We propose a novel, computationally efficient action generation paradigm that preserves the structural benefits of operating on a unit manifold.<n>Our method decomposes the action into a deterministic directional vector and a learnable concentration, enabling efficient between the target direction and uniform noise.<n> Empirically, our method matches or exceeds state-of-the-art methods on standard continuous control benchmarks.
  arXiv  Detail & Related papers  (2025-11-11T13:32:38Z)
• Lookahead Unmasking Elicits Accurate Decoding in Diffusion Language Models [51.12873073612084]
  Masked Diffusion Models (MDMs) as language models generate by iteratively unmasking tokens, yet their performance depends on the inference time order of unmasking.<n>We propose Lookahead Unmasking (LookUM), which addresses these concerns by reformulating sampling as path selection over all possible unmasking orders.<n>LookUM requires only two to three paths to achieve peak performance, demonstrating remarkably efficient path selection.
  arXiv  Detail & Related papers  (2025-11-04T02:37:37Z)
• Enhancing Adversarial Transferability by Balancing Exploration and Exploitation with Gradient-Guided Sampling [82.52485740425321]
  Adrial attacks present a critical challenge to deep neural networks' robustness.<n> transferability of adversarial attacks faces a dilemma between Exploitation (maximizing attack potency) and Exploration (enhancing cross-model generalization)
  arXiv  Detail & Related papers  (2025-11-01T05:43:47Z)
• Adapting to Stochastic and Adversarial Losses in Episodic MDPs with Aggregate Bandit Feedback [61.49239204705301]
  We study online learning in finite-horizon episodic Markov decision processes (MDPs) under the challenging aggregate bandit feedback model.<n>Our results rely on a combination of FTRL over occupancy measures, self-bounding techniques, and new loss estimators inspired by recent advances in online shortest path problems.
  arXiv  Detail & Related papers  (2025-10-20T02:28:08Z)
• READER: Retrieval-Assisted Drafter for Efficient LLM Inference [0.0386965802948046]
  Autoregressive Language Models instantiate a factorized likelihood over token sequences, yet their strictly sequential decoding process imposes an intrinsic lower bound on latency inference.<n>This bottleneck has emerged as a central obstacle to the scalable deployment of large-scale generative models.<n>We present READER, a speculative decoding framework that bypasses the training of the auxiliary draft model.
  arXiv  Detail & Related papers  (2025-08-12T16:47:48Z)
• Improving Black-Box Generative Attacks via Generator Semantic Consistency [51.470649503929344]
  generative attacks produce adversarial examples in a single forward pass at test time.<n>We enforce semantic consistency by aligning the early generator's intermediate features to an EMA teacher.<n>Our approach can be seamlessly integrated into existing generative attacks with consistent improvements in black-box transfer.
  arXiv  Detail & Related papers  (2025-06-23T02:35:09Z)
• Meta-Learning Adversarial Bandits [49.094361442409785]
  We study online learning with bandit feedback across multiple tasks, with the goal of improving average performance across tasks if they are similar according to some natural task-similarity measure. As the first to target the adversarial setting, we design a meta-algorithm that setting-specific guarantees for two important cases: multi-armed bandits (MAB) and bandit optimization (BLO) Our guarantees rely on proving that unregularized follow-the-leader combined with multiplicative weights is enough to online learn a non-smooth and non-B sequence.
  arXiv  Detail & Related papers  (2022-05-27T17:40:32Z)

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.