Academic Intelligence · Curated Daily

Explore the Frontier of Global Academia

AcademicHub aggregates real-time literature from top journals and preprint platforms. Build your personal research radar and let large language models compile cross-disciplinary analysis briefings automatically.

01.
arXiv (quant-ph) 2026-06-25

Exact Leg-Cut Influence Functional and Emergence of Gaussian Entanglement Theory in a Statistical-Dressing Ladder Model

arXiv:2606.25669v1 Announce Type: new Abstract: The emergence of Gaussian effective field theories in low-dimensional quantum systems is traditionally understood through top-down frameworks such as bosonization and Luttinger-liquid theory. However, these approaches typically focus on the long-wavelength degrees of freedom in ways that do not directly track how non-Gaussian lattice-scale correlations are progressively discarded under coarse-graining. In this work, we present an exact lattice formulation from which this phenomenon emerges analytically. We analyze a two-leg hard-core ladder under a leg bipartition, where non-local statistical strings cross the entanglement cut. We construct an exact lattice influence-functional representation showing that the reduced state factorizes strictly into a product-state amplitude and a full-counting-statistics functional. By introducing a commuting linked-cluster superoperator hierarchy that bypasses Baker-Campbell-Hausdorff ordering ambiguities, we prove that the first mixedness-generating sector is strictly density-density in character. Under a specific systematic coarse-graining procedure, we analytically derive the suppression of higher-order corrections, providing a controlled, closed-form framework showing how highly non-Gaussian lattice states evolve toward their quadratic continuum form under coarse-graining. We corroborate these analytical predictions through finite-size exact diagonalization and entanglement-spectrum diagnostics.

02.
arXiv (CS.CL) 2026-06-12

Language Model Circuits Are Sparse in the Neuron Basis

The high-level concepts that a neural network uses to perform computation need not be aligned to individual neurons (Smolensky, 1986). Language model interpretability research has thus turned to techniques which decompose the neuron basis into more interpretable units of model computation, such as sparse autoencoders (SAEs). However, not all neuron-based representations are uninterpretable. For the first time, we empirically show that MLP neurons are as sparse a feature basis as SAEs. We use this finding to develop an end-to-end gradient-based attribution pipeline for circuit tracing on the MLP neuron basis, which surfaces causally effective neurons on a variety of tasks. On a standard subject-verb agreement benchmark (Marks et al., 2025), a circuit of $\approx 10^2$ MLP neurons is enough to control model behaviour. On the multi-hop city-state-capital task from (Lindsey et al., 2025), we find a circuit in which small sets of neurons encode specific latent reasoning steps (e.g. mapping a city to its state), and can be steered to change the model's output. This work thus advances automated interpretability of language models without imposing additional training costs.

03.
arXiv (CS.CL) 2026-06-25

Reclaim Evaluation: A Lossy Memory Is Worse Than an Empty One

Authors:

A language model's memory can be worse than having no memory at all. Give a model a memory that kept a wrong conclusion but dropped the work behind it, and it emits that stale value as a confident answer; give the same model an empty memory and it abstains. Across seven models this direction never reverses, a clean kill condition that none breaks. We call this brittle memory: behavioral, not the near-immediate information bound beneath it; only its magnitude is disposition- and task-dependent, not its direction. We measure it with reclaim evaluation: compress a drifted interaction at a fixed budget, then test whether a correction recovers the known answer, scored against ground truth with no judge. Correctability is bottlenecked by whether the answer-determining source survives, not by capability. A one-line source-first policy (keep the recomputable source, drop the re-derivable conclusion) restores correctability at equal budget where that source is compact and identifiable; a length-matched control rules out added text as the cause. The hand-built oracle reaches 1.00; a one-prompt deployable version reclaims 0.49-0.88. The stake compounds: chained through a memory loop, a single dropped-source error corrupts a growing span of downstream steps and stays uncorrectable, while source-first holds to a bounded budget horizon. The wall and fix replicate across three deployed memory systems and on real dialogue (MultiWOZ), and past the budget where the source no longer fits, the fix fails silently unless the note records completeness. This is a controlled study of a mechanism, not a benchmark: judge-free exact scoring, matched-budget controls, and validators built to come out false. We release the harness, conditions, and validators.

04.
arXiv (CS.AI) 2026-06-16

Stable Menus of Public Goods: AI-Enabled Progress

Authors:

arXiv:2606.16989v1 Announce Type: cross Abstract: Using an open problem from the EC 2025 paper "Stable Menus of Public Goods" as a testbed, we conduct experiments to understand the effectiveness of different AI-for-EconCS research workflows. Specifically, we study three questions: Does providing human intuition in the prompt help? Does automated multi-turn interaction help? And, does an LLM outperform a first-year PhD student? Regarding the first two questions, we provide evidence for the following workflow suggestions: (1) prompting with human intuition can encourage the LLM to have better "taste", (2) multi-turn workflows help when the pipeline encourages "ambitious" steps. Regarding the third question, using an unpublished manuscript written by the paper's senior authors prior to collaborating with the first-year PhD student, we compare the effectiveness of the LLM with that of the first-year PhD student, and find that the LLM is slightly less effective.

05.
PLOS Medicine 2026-06-09

Molecular Tumor Boards clinical impact on patient care and structural features: A systematic review and meta-analysis

Authors:

by Luigi Russo, Erika Giacobini, Nicolò Lentini, Tommaso Osti, Maud Kamal, Stefania Boccia, Roberta Pastorino Background Molecular Tumor Boards (MTBs) bring together multidisciplinary experts to translate genomic data into clinical decisions in oncology, however, their overall clinical impact remains unclear. The aim of this systematic review is to assess the clinical impact of MTB-recommended therapies on patients with cancer outcomes. Methods and findings In this systematic review and meta-analysis, we searched PubMed, Embase, Scopus, and CENTRAL up to July 2025. We included studies of any design, both single-arm studies and studies with a comparator group, that reported the clinical impact of MTBs in patients who received MTB-guided therapy. Meta-analyses were performed separately by study design, using hazard ratios (HRs) for overall survival (OS) and progression-free survival (PFS), relative risks (RRs) for objective response rate (ORR) and disease control rate (DCR), and pooled proportions for PFS ratio ≥1.3. All meta-analyses were conducted using random-effects models based on the inverse variance method. We evaluated the risk of bias using the RoB 2.0 for RCTs and ROBINS-I for non-randomized studies.From 6,846 records, 78 studies (9,195 patients; 4,569 treated per MTB recommendations) were included. MTB-guided therapies were associated with reduced risk of death (HR 0.87; 95% CI [0.76, 1.01]; p = 0.069; I2 = 0.0% in RCTs; 0.62 in retrospective studies) and disease progression (HR 0.73; 95% CI [0.64, 0.84]; p 

06.
arXiv (CS.AI) 2026-06-18

A Convex Route to Thermoelasticity: Learning Internal Energy and Dissipation

arXiv:2603.28707v3 Announce Type: replace-cross Abstract: We present a physics-based neural network framework for the discovery of constitutive models in fully coupled thermomechanics. In contrast to classical formulations based on the Helmholtz energy, we adopt the internal energy and a dissipation potential as primary constitutive functions, expressed in terms of deformation and entropy. This choice avoids the need to enforce mixed convexity–concavity conditions and facilitates a consistent incorporation of thermodynamic principles. In this contribution, we focus on materials without preferred directions or internal variables. While the formulation is posed in terms of entropy, the temperature is treated as the independent observable, and the entropy is inferred internally through the constitutive relation, enabling thermodynamically consistent modeling without requiring entropy data. Thermodynamic admissibility of the networks is guaranteed by construction. The internal energy and dissipation potential are represented by input convex neural networks, ensuring convexity and compliance with the second law. Objectivity, material symmetry, and normalization are embedded directly into the architecture through invariant-based representations and zero-anchored formulations. We demonstrate the performance of the proposed framework on synthetic and experimental datasets, including purely thermal problems and fully coupled thermomechanical responses of soft tissues and filled rubbers. The results show that the learned models accurately capture the underlying constitutive behavior. All code, data, and trained models are made publicly available via https://doi.org/10.5281/zenodo.19248596.

08.
arXiv (CS.CL) 2026-06-16

Rapid Poison: Practical Poisoning Attacks Against the Rapid Response Framework

The Rapid Response (RR) framework, deployed in production systems, including Anthropic's ASL-3 safeguards, continuously improves jailbreak-detection classifiers. When new jailbreaks emerge that bypass these classifiers, Rapid Response generates synthetic variants for training, helping the model generalize from the new attacks and quickly adapt. We reveal that prompt injection can infiltrate this pipeline to deliver poisoned samples into the classifier's training set, enabling two attack objectives: (I) targeted poisoning attacks that create false positives on harmless samples by categorizing them as a jailbreak, with a specific desired feature (e.g., certain formatting, subject, or keyword), (II) concept-based backdoor attacks that induce false negatives on jailbreak inputs, generalizing even to jailbreaks from attack strategies the defender explicitly trained against, when the backdoor trigger is present. Importantly, our threat model restricts adversaries to modifying only jailbreak samples (not benign data or labels), a constraint unexplored by prior work that makes the second objective particularly challenging. We address this with Omission Attack, which exploits a new phenomenon: when training on concept-absent unsafe samples, the classifier misassociates that concept's presence with the safe label. Both attacks cause substantial and in some cases near-complete label flipping at only a 1% poisoning rate, achieving up to 100% false positive rates and up to 96% false negative rates.

10.
arXiv (CS.CV) 2026-06-24

DiffusionBench: On Holistic Evaluation of Diffusion Transformers

Diffusion transformer (DiT) research on image generation has converged to a single evaluation setup: class-conditional generation on ImageNet. While methods improve the FID and related metrics, it is increasingly unclear whether they reflect real progress in generative modeling. The natural alternative, i.e., text-to-image (T2I) generation, is perceived as too costly or inconvenient to train and evaluate and is often skipped. We argue that this perception no longer holds. We introduce NanoGen, a unified DiT training and evaluation framework. NanoGen matches state-of-the-art DiT baselines on ImageNet and, with 12 lines of configuration change, also trains competitive text-to-image models. It currently supports RAE, VAE, pixel-space, and MeanFlow diffusion methods under both ImageNet and T2I setups. Under NanoGen, training T2I requires comparable compute to ImageNet. After training 21 latent diffusion models with NanoGen, we observe that method ranking shows no strong correlation between ImageNet and T2I generation: Pearson correlation is between -0.377 and -0.580 across three metrics. This suggests that a method which improves class-conditional ImageNet FID may show no corresponding improvement on T2I, clearly indicating the necessity of evaluating DiTs on both tasks. To this end, we summarize ImageNet and text-to-image results, which yields DiffusionBench, a holistic benchmark for DiT research. We recommend reporting DiffusionBench in place of ImageNet alone: methods that improve DiffusionBench are more likely to reflect broader progress.

11.
arXiv (CS.CL) 2026-06-16

PathRouter: Aligning Rewards with Retrieval Quality in Agentic Graph Retrieval-Augmented Generation

Agentic GraphRAG trains language-model agents to iteratively retrieve and reason over graph-structured evidence, enabling more accurate and context-aware decision-making by efficiently navigating complex information networks. However, outcome-only reinforcement learning suffers from answer-path reward aliasing, where correct answers may come from shortcuts rather than useful evidence paths. It also exhibits search-update ambiguity, as scalar trajectory-level feedback does not indicate which retrieval actions to adjust. To mitigate these shortcomings, we present PathRouter, a path-aware training framework for agentic GraphRAG. PathRouter jointly evaluates each trajectory along answer correctness and evidence-path overlap, yielding four trajectory categories with differentiated GRPO advantage scaling that suppresses shortcut reinforcement while preserving evidence-seeking behavior. For evidence-poor trajectories, a frozen gold-evidence teacher provides token-level KL guidance on reasoning and search-query tokens, excluding answer tokens to avoid direct response imitation. Experiments on six QA benchmarks across three model sizes show that PathRouter consistently improves answer F1 and evidence-path overlap, achieving average F1 gains of 3.1 on 3B and 4.9 on 7B models compared to a strong baseline.

12.
arXiv (CS.CL) 2026-06-25

Evaluating LLMs on Real-World Software Performance Optimization

Software performance optimization is a notoriously complex and manual task. Despite the growing use of Large Language Models (LLMs) for code refinement, we still lack benchmarks that capture how optimization actually happens in real-world codebases. Existing frameworks often oversimplify the problem by focusing on isolated functions or a single performance metric, missing the critical trade-offs between execution time and memory footprint, the inherent noise of the measurement environment, and the variability introduced by different input data and execution conditions. We address this by introducing SWE-Pro, a repository-level benchmark derived from 102 expert-written optimizations from open-source projects. Unlike previous benchmarks, SWE-Pro pairs each task with parameterized tests to evaluate runtime, peak memory, and Time-Weighted Memory Usage (TWMU) across varying input data and execution conditions under noise-aware measurement conditions. Our evaluation shows that current LLMs struggle significantly: runtime gains are negligible, and memory optimizations are nearly non-existent. This stands in sharp contrast to expert implementations, which achieve an aggregate speedup of 15.5x and peak memory reduction of 171.3x over benchmark tasks. Expert-written improvements are observed in 91.2% of tasks for runtime and 65.7% for peak memory. Our findings expose a substantial gap between current LLM capabilities and the demands of expert-level engineering.

13.
arXiv (CS.CL) 2026-06-12

Operads for compositional reasoning in LLMs

Question decomposition, i.e. breaking a complex query into simpler sub-queries whose answers are composed to produce a final answer, is a widely used strategy for improving LLM reasoning, yet it currently lacks a rigorous mathematical foundation. In this paper, we propose operads, mathematical structures that model many-in, one-out operations and compositions thereof, as a natural framework for describing question decomposition. We define the questions operad $Q$, in which operations correspond to question templates and composition corresponds to substitution of sub-answers, and show how QA models can be interpreted as algebras over $Q$. Beyond reframing existing practice, this operadic perspective points toward new methods, in particular a notion of operadic consistency, which measures whether a QA model's answers agree across the partial collapses of a question decomposition tree. Empirical evaluation of operadic consistency is reported in our companion paper (Bottman, Liu, and Richardson, 2026), which finds it strongly correlated with accuracy across twelve LLMs and four multi-hop QA datasets and outperforming standard temperature-based self-consistency baselines. We argue that operads are the natural mathematical home for question decomposition, and that invariants such as operadic consistency open new directions for analyzing and improving the reliability of multi-step reasoning.

14.
arXiv (CS.LG) 2026-06-16

GRASP: Gradient-Aligned Sequential Parameter Transfer for Memory-Efficient Multi-Source Learning

arXiv:2606.14900v1 Announce Type: new Abstract: Multi-source transfer learning faces a fundamental scalability bottleneck: existing approaches require either loading all K source models into memory simultaneously during parameter fusion, requiring O(K) memory, or deploying all models at inference time, making production deployment infeasible. We propose GRASP (Gradient-Aligned Sequential Parameter Transfer), which achieves superior knowledge integration while maintaining O(1) memory consumption through three key innovations: (1) sequential processing that merges one source at a time into an evolving target model, (2) parameter-wise gradient alignment that selectively transfers only parameters whose optimization directions align with the target domain, avoiding negative transfer, and (3) iterative fine-tuning that adapts transferred knowledge before integrating the next source. Extensive experiments across three continual learning benchmarks (Yearbook, CLEAR-10, CLEAR-100) spanning 10 to 108-year temporal distribution shifts and four architectures (1.3M to 25.6M parameters) demonstrate that GRASP achieves 93.5% mean accuracy over all datasets and architectures compared to ensemble method's 71.7% accuracy while requiring only constant memory versus K models for standard multi-source fusion. Critically, GRASP's sequential previously merged models and scales to arbitrarily many sources without memory growth, making it uniquely suitable for resource-constrained deployment and continually evolving source domains.

15.
bioRxiv (Bioinfo) 2026-06-11

DLDN-Bench: A Benchmark Framework for Deep Learning de Novo Peptide Sequencing in Proteomics

De novo peptide sequencing is an essential approach for analyzing mass spectrometry data because it enables the identification of novel peptides without relying on protein sequence databases. Recent advances in deep learning have substantially improved the performance of de novo sequencing methods, but the rapid emergence of new models has led to heterogeneous evaluation practices and limited comparability. To address this, we introduce DLDN-Bench, a benchmark framework including a set of benchmark datasets derived from human muscle biopsy mass spectrometry data retrieved from PRIDE and annotated through consensus across multiple widely used database search engines. Using these datasets, we systematically benchmark recent deep learning-based de novo sequencing tools alongside traditional approaches. Performance is assessed using established metrics, including precision and coverage relative to a pseudo-ground truth defined by cross-engine agreement. To demonstrate the utility of DLDN-Bench, we benchmark four recent deep learning models and make all results publicly available. This benchmark framework provides a standardized basis for comparing state-of-the-art methods and offers an extensible resource for evaluating future tools in de novo peptide sequencing.

16.
arXiv (CS.AI) 2026-06-25

Logit Distance Bounds Representational Similarity

arXiv:2602.15438v3 Announce Type: replace-cross Abstract: For a broad family of discriminative models that includes autoregressive language models, identifiability results imply that if two models induce the same conditional distributions, then their internal representations are equal up to an invertible linear transformation. We ask whether an analogous conclusion holds approximately when the distributions are close instead of equal. Building on the observation of Nielsen et al. (2025) that closeness in KL divergence need not imply high linear representational similarity, we study a distributional distance based on logit differences and show that closeness in this distance does yield linear similarity guarantees. Specifically, we define a representational dissimilarity measure based on the models' identifiability class and prove that it is bounded by the logit distance. We further show that, when model probabilities are bounded away from zero, KL divergence upper-bounds logit distance; yet the resulting bound fails to provide nontrivial control in practice. As a consequence, KL-based distillation can match a teacher's predictions while failing to preserve linear representational properties, such as linear-probe recoverability of human-interpretable concepts. In distillation experiments on synthetic and image datasets, logit-distance distillation yields students with higher linear representational similarity and better preservation of the teacher's linearly recoverable concepts.

17.
arXiv (CS.LG) 2026-06-19

Reinforcement Twinning for Hybrid Control of Flapping-Wing Drones

arXiv:2505.18201v2 Announce Type: replace-cross Abstract: Controlling flapping-wing drones requires controllers that handle time-varying, nonlinear, underactuated dynamics from incomplete, noisy sensor data. Recent advances in artificial intelligence (AI), particularly reinforcement learning (RL), have opened new perspectives for addressing such complex control problems through data-driven policy optimization from interaction with the environment. Yet purely data-driven methods are sample-inefficient, demanding extensive, sometimes unsafe exploration, especially without guiding physical models. This motivates hybrid AI-physics frameworks. This article proposes a hybrid model-free/model-based flight-control approach using the reinforcement twinning algorithm. The model-based (MB) component uses an adjoint formulation and an adaptive digital twin continuously identified from live trajectories; the model-free (MF) component uses RL. The two agents share knowledge via transfer learning, imitation learning, and shared experience between the real environment and the digital twin, coordinated by a policy referee that selects which agent acts in reality based on digital-twin performance and a real-to-virtual consistency ratio. The framework is evaluated for the longitudinal control of a flapping-wing drone, modelled as a nonlinear time-varying system driven by quasi-steady aerodynamic forces. The hybrid strategy is tested under three adaptive-model initializations: (1) offline identification from existing data, (2) random initialization with fully online identification, and (3) offline pre-training with biased parameters followed by online adaptation. In all cases, the hybrid framework improves performance, robustness, and sample efficiency over purely model-free and purely model-based approaches.

18.
arXiv (CS.CL) 2026-06-11

When Probing Accuracy Saturates, Fragility Resolves: A Complementary Metric for LLM Pre-Training Analysis

Standard linear probing declares a property "encoded" when a classifier on hidden states achieves high accuracy. The protocol works well on a snapshot but breaks across pre-training: probe accuracy saturates within the first few thousand steps, leaving most of training invisible to the instrument. We introduce fragility, a complementary per-layer metric defined as the activation-noise level at which probe accuracy collapses. Fragility is sensitive to both the margin of separability and the redundancy of representation, both of which keep evolving long after accuracy plateaus. Applied to open-checkpoint language models, fragility recovers structure that accuracy alone cannot see. Moralized representations emerge along a lexical $\to$ compositional gradient: lexical moral detection first, compositional moral encoding later. Because probe accuracy on its own tracks how lexically separable a dataset is, we establish the compositional encoding directly, by showing it transfers across construction types that share no contrast tokens. A layer-depth robustness gradient develops monotonically across training while accuracy stays flat. And matched fine-tuning corpora that produce identical probing accuracy leave distinct fragility fingerprints, showing that data curation reshapes probe robustness without changing probe accuracy. In every comparison we test, where probing accuracy returns a flat answer, fragility returns a structured one.

19.
arXiv (CS.AI) 2026-06-18

From Specification to Execution: AI Assisted Scientific Workflow Management

arXiv:2606.18425v1 Announce Type: cross Abstract: Scientific workflow management systems (WMS) support scalable and reproducible execution of complex pipelines, but workflow design, implementation, and debugging remain largely manual and require significant expertise. Recent approaches using large language models (LLMs) show promise for workflow generation from natural language, but often rely on direct code synthesis, which limits transparency, reproducibility, and integration with workflow systems. We present an AI-assisted approach to scientific workflow management that combines specification-driven workflow generation, automated debugging, and distributed execution. The method introduces a structured specification phase that separates workflow intent, design, and implementation, allowing validation prior to code generation. We also develop an LLM-based debugging agent that diagnoses and resolves failures across multiple system layers. To support distributed execution and user interaction, we integrate Pegasus, a widely used WMS, with a Model Context Protocol (MCP) layer, providing a unified interface for workflow submission, monitoring, and control. We evaluate the approach using a federated learning workflow for medical imaging, chosen for its parallel, iterative, and dependency-intensive structure. The system generated and executed large-scale workflows with thousands of jobs, reduced debugging effort, and allowed non-expert users to construct workflows with expert-level design patterns. These results indicate that end-to-end AI-assisted workflow generation and execution is feasible, and point toward AI-driven platforms for managing the scientific workflow lifecycle.

20.
arXiv (CS.CL) 2026-06-19

The ACUTE Protocol: Operationalizing Language Model Activations for Better Calibration, Utility, and Trust

As language models improve and become increasingly deployed to solve a variety of tasks, trustworthiness becomes essential. Calibration is a good proxy for trust: well-calibrated confidence estimates help inform the risk versus reward tradeoff when trusting a specific model output. Unfortunately, even as models improve, they remain poorly calibrated, often biasing towards overconfidence. Additionally, calibration can be gamed: a policy that always predicts the base rate is perfectly calibrated, but completely uninformative. To resolve this, we develop a new metric, expected utility renormalized by the oracle (EURO), that balances calibration and informativeness. We also propose a general-purpose activation-based confidence, utility, and trust estimation protocol (ACUTE) to appropriately adjudicate uncertainty. The ACUTE protocol provides flexible, sample-efficient, and compute-efficient confidence estimators for 3 tasks including multiple choice question answering, tool-calling, and scientific document summarization across 6 models from 4 model families. ACUTE outperforms strong baselines on EURO, while maintaining low calibration error. Taken together, our work shows that equipping LLMs with the ACUTE protocol can improve calibration, utility, and trustworthiness in numerous settings.

21.
arXiv (CS.CL) 2026-06-24

BehaviorBench: Benchmarking Foundation Models for Behavioral Science Tasks

Foundation models have been increasingly applied to behavioral science domains such as psychology, sociology, and economics. While these models show promise in individual tasks such as survey response prediction and human-subject experiment simulation, there remains no systematic understanding of how well they perform across diverse behavioral science tasks, contexts, and populations. We introduce BehaviorBench, a comprehensive benchmark that evaluates foundation models along four core capabilities: (1) behavior prediction and simulation, (2) strategic decision-making, (3) subject-trait inference, and (4) behavioral knowledge application. Crucially, BehaviorBench evaluates model outputs at both the individual and distributional levels, capturing not only per-subject accuracy but also population-level alignment, an essential requirement for behavioral validity. Leveraging the tasks in BehaviorBench, we further develop Be.FM-1.5, extending the Be.FM family of behavioral foundation models fine-tuned on behavioral data. Our results reveal a considerable gap: proprietary general-purpose models excel at individual-level prediction and knowledge-intensive tasks, whereas behavioral foundation models, fine-tuned on behavioral data, achieve substantially stronger distributional alignment. Notably, Be.FM-1.5 leads on distributional metrics and remains competitive on individual-level metrics, suggesting that proper behavioral adaptation can close the gap. Our results highlight the importance of distributional evaluation, establish BehaviorBench as a foundation for developing and assessing behaviorally aligned AI systems, and demonstrate Be.FM-1.5's potential for a broad range of behavioral science studies. Our BehaviorBench and Be.FM-1.5 models can be accessed via https://umich-foreseer.github.io/behaviorbench/.

22.
arXiv (CS.CL) 2026-06-17

Learning task-specific subspaces via interventional post-training of speech foundation models

Speech foundation models, pre-trained on large corpora of unlabelled speech data, produce general-purpose representations which are useful across tasks. However, these representations encode information about salient speech variables in a distributed manner, while downstream speech tasks rely on only some of this variability. In this work, we propose a post-training refinement approach using interventional contrastive learning. By leveraging an interventional dataset and multi-part contrastive loss, we learn a transformation from the entangled representation space of speech foundation models into separate content and speaker subspaces. We evaluate the learnt representations on speaker verification and keyword spotting tasks, showing improved out-of-domain speaker verification performance and evidence that speaker and content information are separated across the learned subspaces.

23.
arXiv (CS.CL) 2026-06-11

Self-Attention as Transport: Limits of Symmetric Spectral Diagnostics

When a language model processes a hallucinated response, its attention routing tends to fail in one of two shapes: over-concentrating on a narrow set of positions, or spreading so diffusely that relevance is diluted, and the shape of the failure carries diagnostic signal. We study these shapes as a diagnostic characterization, computed from attention matrices under forced scoring of benchmark-labeled responses rather than during live generation. A widely used family of spectral methods analyzes the symmetric component of the degree-normalized attention operator, which governs transport capacity; we prove that every transpose-invariant spectral diagnostic of this operator is structurally orientation-blind (it cannot distinguish an operator from its transpose, and therefore cannot detect information-flow direction), with a converse to the blindness theorem bounding any Lipschitz diagnostic's transpose sensitivity by the asymmetry coefficient $G$. Pairing this with a closed-form bipartite-Cheeger landscape for canonical causal architectures, we show that uniform causal attention satisfies an $n$-independent floor $\phi \ge 1/5$, while window attention pierces the floor as $O(w/n)$; failure modes are shape-different, not just value-different. This floor is an idealized-architecture benchmark, not an empirical attractor: the fraction of real attention heads that pierce it is itself an architectural signature. The resulting two-axis diagnostic ($\phi$ for capacity, $G$ for direction) yields a falsifiable polarity prediction: bottleneck- and diffuse-dominated benchmarks should exhibit opposite polarity. Under length-controlled evaluation, transport features retain interpretable signal (0.62-0.84 LC-AUROC) across the tested decoder-only, encoder-only, and encoder-decoder models, with polarity reversing as predicted between HaluEval and MedHallu.

24.
arXiv (CS.CL) 2026-06-17

Smarter edits? Post-editing with error highlights and translation suggestions

As MT quality increases, interest in enhanced post-editing features such as QE-derived error highlights is growing, yet evidence for their usefulness remains limited. In this work, we explore the usefulness of LLM-derived error highlights and correction suggestions based on automatic post-editing (APE). We conduct a study where professional translators (En-Nl) post-edit translations using APE error highlights and correction suggestions and compare productivity, quality and user experience to regular PE and PE with QE-derived highlights. While no condition yielded productivity or quality gains compared to regular PE, APE highlights were better received than QE-derived highlights, and correction suggestions improved overall user experience.

25.
arXiv (CS.CL) 2026-06-11

ALIGNBEAM : Inference-Time Alignment Transfer via Cross-Vocabulary Logit Mixing

Domain fine-tuning degrades the safety of large language models: fine-tuned specialists readily comply with harmful prompts framed in domain language. Existing inference-time defenses that mix logits from a safe anchor model require both models to share a vocabulary, which rules them out for the cross-family specialists where safety is most degraded. We present ALIGNBEAM, a training-free method that lifts this restriction by translating anchor logits into the target model's vocabulary token-by-token at each decoding step; a small LLM judge then selects the safest among K candidate continuations. No weights are changed, and the safety-utility trade-off can be tuned at deployment without retraining. Across both cross-vocabulary and same-vocabulary evaluation pairs, ALIGNBEAM substantially raises refusal on adversarial benchmarks while keeping task accuracy and inference overhead within practical bounds. The results show that safety alignment can be transferred between model families at inference time, without touching either model's weights.