Reconstructing audio from brain activity remains challenging. Existing brain-to-audio pipelines typically map neural responses to predefined stimulus representations before synthesis. While recent work in the visual domain has explored direct conditioning on brain activity, this approach has not yet been extended to audio diffusion models. We introduce MAAS (Multi-Area Auditory Synthesis), an end-to-end framework that conditions a pretrained audio diffusion transformer on ROI-wise fMRI responses without first supervising a decoder to predict an external stimulus representation. MAAS uses region-specific projections and learned brain tokens to condition Stable Audio through a dedicated cross-attention pathway. On a 7T fMRI dataset, we demonstrate consistent improvements over baselines, including two-stage fMRI-to-latent conditioning, across both acoustic and semantic evaluation metrics. In addition, we introduce a fidelity metric grounded in a computational model of auditory neural processing, which allows us to quantify the contribution of individual auditory cortical regions to reconstructed audio representations.
The traveling thief problem (TTP) asks a thief to choose a tour and a packing together, and the benchmark of 9,720 instances of Polyakovskiy et al. (2014) is the standard test set for comparing TTP heuristics. We show that four of these instances are not TTP instances at all. Their files give a renting rate of 0.00, so the tour has no effect on the objective and each instance reduces to a 0-1 knapsack problem. One of them belongs to the 60 categorized instances on which new methods are usually compared, and on it five of nine published methods report exactly the same value. The cause is the format of the files, which keep the renting rate with two decimals, while the rule of the generator gives a rate close to 0.003 on these instances. The same rounding may shift the rate of 949 other instances by more than 1%. Since the tour and the packing used by the generator were not published, the original rates cannot be recovered. We therefore define new rates with the same rule, applied to a fully specified reference solution that we publish, and we show with a published heuristic that the corrected instances behave as TTP instances again. We also give reference results on the corrected instances and recommendations for future studies. Finally, we present ttplib, a new open-source Python library that loads any of the 9,720 instances, including the corrected ones, without downloading the whole benchmark.
Large language models (LLMs) are increasingly used to translate technical explainable AI (XAI) artefacts such as SHAP into explanatory narratives (XAINs) for lay users. In high-risk domains, these explanations are often the main resource through which individuals can understand and contest adverse decisions. Yet LLMs are known to reproduce social bias, even from proxy variables alone, and fairness in XAINs remains unstudied. Because XAINs are bound to a ground-truth artefact, fairness concerns not only what a narrative says and how, but also whether it is equally faithful across social groups. We introduce fairness in XAINs as a new field of research and propose a framework that generates narratives under three conditions (protected attribute disclosed, withheld, or counterfactually swapped), allowing five comparisons that pinpoint the cause of disparities. We apply it to adversely classified instances from four datasets in credit scoring, education and employment, generating narratives with five LLMs, and evaluate faithfulness, feature coverage and sentiment. We find significant disparities across all datasets, metric families and models, arising in two ways. Faithfulness disparities result from feature values: LLMs make plausible but false statements, triggered by values that are unequally distributed across social groups. Feature coverage and sentiment disparities stem from the protected attribute itself: once it is disclosed, LLMs give less explanation to, or frame more negatively the narratives of, groups that have historically been discriminated against or are stereotypically associated with worse outcomes. A more faithful model removes the faithfulness disparities, and withholding protected attributes mostly removes the others, but neither suffices on its own. Fairness in XAINs should therefore be audited separately from the predictive model and its XAI artefact.
Optimizing molecules under small evaluation budgets requires deciding both where in chemical space to search and which candidate within that region to evaluate. We introduce Trust Region Adaptive Chemical Exploration (TRACE), a framework for open-ended molecular Bayesian optimization (BO), where candidates are generated on demand rather than selected from a fixed library. TRACE combines an objective-free graph-perturbation mechanism with a fixed, chemistry-aware Gaussian process (GP) surrogate using fragprints, a representation combining molecular fingerprints and fragment descriptors. Rather than learning the search geometry itself, TRACE makes the scale of chemical exploration explicit through a Tanimoto-similarity threshold controlling chemical extent and a cardinality budget controlling candidate-search breadth. A simple success/failure controller expands or contracts both quantities during optimization, while the GP and acquisition function select which admissible molecules to evaluate. Across six challenging molecular optimization benchmarks under matched 1,000-evaluation budgets, TRACE achieves the strongest overall sample efficiency as measured by the best molecule found as evaluations accumulate. Crossed ablations on two objectives further show that adaptive region control contributes beyond the proposal mechanism, while its benefit depends on the proposal distribution. We further apply TRACE to the stability of iron phthalocyanine (FePc), a non-precious molecular electrocatalyst whose degradation can involve loss of the Fe center. Using a first-principles density functional theory (DFT) workflow designed for accurate energy evaluation, TRACE increases the computed thermodynamic resistance to Fe dissolution from 0.42 to 2.59 eV over 80 designed evaluations.
Reinforcement learning with verifiable rewards ordinarily spreads a single advantage over a whole trajectory, a granularity too coarse to isolate the few decisions on which success in a long, multi-turn interaction actually turns. Privileged self-distillation supplies denser supervision, but it leaves unanswered how a set of local teacher-student discrepancies should be converted into credit along a sequence. We introduce AgentOPSD, a critic-free scheme that assigns credit recursively, one interaction turn at a time. Inside a turn, the token-level log-probability gaps between a skill-conditioned self-teacher and the student are summed, and that sum acts as evidence in a recursive Bayesian update carried out in log-odds space; the resulting revisions reallocate the outcome-derived advantage across turns through a multiplier that is bounded and preserves its sign. With Qwen2.5-3B and Qwen2.5-7B as policy backbones on ALFWorld, the method improves consistently on GRPO and on self-distillation baselines, and attains 89.1% success at the 7B scale. Ablations attribute the gain to aggregating evidence at turn boundaries and to making the belief revision depend on history, rather than to the local gap by itself.
Likelihood-based methods are widely used to learn the parameters of partially observable Markov decision processes (POMDPs). While finite-sample guarantees for policy suboptimality exist for certain POMDP subclasses, analogous guarantees for parameter recovery remain less well understood. For any candidate POMDP parameter, we bound its parameter recovery error in terms of its likelihood gap to the ground truth, the structural conditioning of the observation and transition kernels, and the behavior policy used to collect data. The bound applies in both the undercomplete and overcomplete regimes, with the latter following naturally from the undercomplete analysis. Because the result applies to any sufficiently good candidate, it yields finite-sample guarantees for the maximum-likelihood estimator and any procedure producing a parameter with controlled likelihood gap. This is particularly useful for near-optimal learning in infinite-horizon POMDPs, where existing approaches largely rely on model-based planning with accurate transition and observation kernel estimates. The bound separates the contributions of the behavior policy and model, showing how data-collection choices, including action coverage, and their effects on mixing can accelerate parameter learning.
As vision-language models (VLMs) move toward interpreting medical images directly, their outputs increasingly include not just a label but a stated confidence and a written justification. Yet the field still predominantly evaluates VLMs using just accuracy. We evaluate the coherence between answer, confidence, and explanation across four VLMs on breast density and BI-RADS categorization, using two mammography datasets. We perform controlled perturbations of the question, multiple choice option orders, and the images and assess changes in the three outputs. Our results suggest confidence does not decrease even as the mammography image is changed. Shuffling the options also changes most VLMs' answers. Notably, there are cases where a model's own stated reasoning contradicts its final answer, a distinction accuracy alone cannot capture but one that governs whether an AI-generated answer can be trusted.
Many tasks compose transformations by a known rule: chained rotations in robotics and navigation. Learned models relearn it, compounding error; classical estimators hardcode it and miss hardware deviations. We impose the known structure and learn only the remainder: an exact non-commutative group law sits in the node of a self-contained layer, its edges learn local increments, and a zero-initialized correction psi learns unknown deviations of the law. A pre-training criterion predicts exact representability. On fixed-angle SU(2) composition, a 54-parameter dense edge is the most accurate dense model, at 27x fewer parameters than the strongest fixed-matrix baseline, and lies 0.05 deg above the Bayes floor. On real robot arms, an isolated 306-parameter psi reaches 1.12-1.14 mm versus 1.29-1.33 mm for Gauss-Newton at N >= 60, and the recipe transfers across three vendors and two protocols. The recipe matches classical fits where the data obey the law and loses where the imposed law is not the task's.
Vision language models (VLMs) encode substantial expert knowledge within their weights, yet their direct application to specialised classification domains such as benthic imagery remains limited. Meanwhile, vision transformers trained on these tasks rely predominantly on low-level visual latents (colours, textures, and shapes) that lack the higher-level reasoning humans employ for identification. We propose Guided Hierarchical Multi-Label (GHiML) learning, a pipeline that extracts conceptual latent variables from a VLM conditioned on training data, generates sample-relevant captions grounded in those variables, and fuses the resulting textual embeddings with image embeddings via cross-attention and a per-node gated router. We also introduce the inverted HML score ($d_{\mathrm{inv}}$), a similarity score that evaluates hierarchical multi-label predictions through graph traversal distance and a cardinality penalty, providing an interpretable alternative to existing measures, which are insensitive to realistic degradations in sparse, deep hierarchies. Experiments on BenthicNet, FathomNet and iNaturalist show that the router improves macro $F_1$ over the vision-only baseline by up to 6.7 pp and $d_{\mathrm{inv}}$ by up to 21.1 pp on in-domain benthic data, with gains persisting under distribution shift. Ablations locate the contribution of the text itself: on the multi-label biotic hierarchies, removing the per-sample text signal degrades $d_{\mathrm{inv}}$ while leaving macro $F_1$ unchanged, consistent with text acting through prediction cardinality rather than node discrimination, whereas iNaturalist, whose annotations are single-label, loses 18.7% of macro $F_1$. Our results support the hypothesis that VLMs contain exploitable specialised knowledge that, when appropriately elicited, can augment classification in domains where direct visual learning is insufficient.
Video world models conditioned on an input video make it possible to explore real, recorded dynamic scenes, enabling re-shooting of footage, free-viewpoint replay, and real-to-sim environments for embodied agents. Such uses require the generated rollout to stay faithful to the recorded world, yet existing benchmarks start from a single image or a text prompt, and video-conditioned generation is evaluated mainly on perceptual quality and camera following. These evaluations do not directly assess whether exploration preserves subject identity, event progression, and scene consistency upon revisiting a location. We introduce WanderBench, a benchmark for exploring dynamic worlds from video. It comprises 745 single-shot source videos of real and AI-generated content in three difficulty tiers (3-25 s), paired with camera routes given as keyboard actions, and evaluates six dimensions with 24 human-validated sub-metrics. Benchmarking eight models, we find that as clips grow longer, models keep the static scene and visual quality but lose camera control and the source dynamics, and that none keeps revisited places consistent. We further propose WanderLand, a training-free framework that conditions each frame of a frozen model on the same-instant source and, on revisits, on its own earlier static content. WanderLand improves five models on every dimension, by up to 8.3 points in source-dynamics fidelity.
We propose SciMARS, a framework for recursive task-local adaptation of multi-agent harnesses in closed-loop scientific discovery. Multi-agent systems coordinate analysis, experimental design, and verification as scientific evidence evolves. However, two problems limit their effectiveness: (i) without shared memory, agents lose observations and repeat work; (ii) fixed role procedures, handoffs, and memory policies perpetuate mistakes despite retained evidence. We address these problems with a shared evidence graph and separate component editors. SciMARS adapts roles and coordination; SciMARS + Memory adds a third editor for how agents record, retrieve, and receive evidence. Candidate probes guide composition and selection of the next harness. We evaluate scientific discovery through experimental-data replay and a digital twin. SciMARS + Memory resolves 12/18 GPT-6-Astra and 2/18 GLM-5.3 attempts on Autonomous Science Bench, compared with 8/18 and 1/18 for the base multi-agent team. On ScienceAgentBench and AppWorld, it solves 14/38 and 11/20 tasks, compared with 11/38 and 5/20 for the base team. These results support combining shared memory with harness adaptation across scientific and interactive tasks.
We prove $F_{\mathrm{weighted}}(2,2,3)<11/6$ for minimum-Frobenius-norm least-squares regression with two input and two output coordinates, improving the upper bound $15/8$. More precisely, a universal $\varepsilon_0>0$ satisfies $\mathcal T_3\le(5/6-\varepsilon_0)E$ for every finite residual frame. The proof combines an exact scalar certificate, rigidity of equality in planar selection, and a compactness argument that controls disappearing weights. Together with the known lower bound, this gives $13/8\le F_{\mathrm{weighted}}(2,2,3)<11/6$. No numerical value of $\varepsilon_0$ is claimed.
Alternating minimization is a standard tool for optimization problems with two coupled variable blocks. Its cost is dominated by the inner subproblem solvers, which rarely have a closed-form solution. In this paper, we propose to learn an inexact alternating scheme in which neural networks approximate the subproblem solution at each iteration. For convex problems with quadratic couplings between the variable blocks that satisfy a quadratic growth condition, we derive a worst-case bound on the optimality gap that exposes the per-iteration error that each network is trained to minimize in a greedy fashion. Combining the bound with a PAC-Bayes argument, we obtain a fixed-horizon, high-probability bound on the expected optimality gap over the distribution of problem instances. We apply the learned scheme to robust matrix completion and low-dose CT reconstruction with dictionary regularization where we achieve a 5x-13x speedup over the inexact first-order baseline at matched accuracy. A brief end-to-end fine-tuning stage extends this up to 30x and outperforms other learned schemes. The code is available at https://github.com/paulhausner/learned-alternating-minimization.
Multilingual generation requires large language models (LLMs) to determine both what to say and in what language to say it, yet these two commitments are difficult to disentangle from model outputs alone. We use translation as a controlled probe: for multi-parallel segments that a model translates well in one language direction but poorly in another, content is held fixed while language varies. Using activation patching, we write residual-stream states from successful runs into failing ones and compare content-matched states with content-independent controls, allowing us to trace content-specific and language-related signals across layers. Across three LLMs of different scales and architectures, we find a consistent three-stage pattern of multilingual generation. Intermediate layers first express a default English signal; content information becomes causally available one layer later; and the final target language is resolved only at later layers. The ordering of the English default before content also appears in a reading-comprehension task. Our sparse autoencoder analyses reveal a further dissociation: the intermediate language signal is only weakly represented by sparse features, whereas a later language distinction is clearly sparse, but resists intervention at the patched position. These results provide a mechanistic account of how language selection and content information emerge and interact during multilingual generation.
Self-recurrence represents an important component for temporal processing in neural networks, both artificial and biological, where it appears naturally in the form of leaky integration. However, due to the complex temporal dependencies of the gradient, an efficient and effective method of online training is still lacking. Here we propose K-prop, a novel online learning method for leaky neural networks. It dramatically reduces the computational and memory cost of real time recurrent learning (RTRL) by propagating kernel gains. K-prop also outperforms existing RTRL approximations and achieves performance comparable to backpropagation through time in sequence learning, classification, and reinforcement learning tasks. We further analyze how its sole approximation affects gradient quality and identify conditions under which K-prop remains accurate and effective.
Prior beliefs about whether a virtual human (VH) is controlled by a human or a computer can shape how users interpret and respond to its behaviour. As emerging hybrid VH systems allow control to transfer between human operators and computer programs during an ongoing interaction, it remains unclear whether perceived agency updates with the actual controller or continues to reflect users' prior beliefs and first impressions. We investigated this question in a within-subject VR study with 24 participants across four conditions: always computer-controlled, computer-to-human, human-to-computer, and always human-controlled. We designed an interaction task in which participants acted as interviewers, asking the VHs a fixed set of questions. Before each interaction, participants were informed whether the VH was human- or computer-controlled, matching the actual controller at the beginning, but were unaware of possible control transitions. We found that perceived agency was strongly influenced by the controller presented at the beginning of the interaction, with no evidence of updating following the control transitions. Social evaluations including likeability and co-presence were generally higher in conditions where human control occurred later and covered a greater proportion of the interaction. The results suggest that prior beliefs and first impressions continued to shape perceived agency, whereas social evaluations were more sensitive to the VH's actual behaviour over the interaction.
This work proposes WaveGen, a fully end-to-end diffusion-based text-to-waveform framework based on Spectral Forcing, which jointly learns an internal spectral trajectory and the target waveform trajectory within a single model, enabling direct end-to-end text-to-waveform generation. Unlike recent TTS pipelines, WaveGen does not rely on separately trained or externally decoded intermediate representations, such as neural audio codecs, Mel-spectrogram vocoders, or VAE-based latent models. In addition, WaveGen performs internal text-speech alignment within a single model, eliminating the need for external alignment modules such as duration predictors. To preserve a truly end-to-end formulation, our framework further avoids dependence on self-supervised representations, such as BERT-based text embeddings and Wav2Vec 2.0-based semantic representations, as well as semantic distillation methods for training optimization. To achieve this, we carefully design Spectral Forcing, a model architecture with disentangled diffusion heads, and a training objective based on a multi-scale spectral $v$-loss. We further validate the scalability of end-to-end TTS models by scaling the model size from 0.1B to 0.6B parameters. Using a single architecture, open-source data, and one-stage training, without external modules such as duration predictor or REPA, WaveGen demonstrates fully end-to-end waveform generation achieving state-of-the-art performance on the Emilia training dataset. In particular, WaveGen achieves a WER of 1.47 and a SIM of 0.75 on the Seed-en benchmark. Source code and checkpoints are available at https://github.com/sh-lee-prml/WaveGen
Vision-language models (VLMs) are increasingly used as automated judges for image-editing benchmarks, yet their reliability is poorly understood. Auditing these judges is difficult because real edited images lack controlled ground truth and entangle the requested changes with model artefacts, unintended side effects, and ambiguous degrees of success. In this paper, we introduce EditJudge-Bench, a diagnostic benchmark of 1,382 synthetic before/after image pairs rendered in Blender, with edits parameterised across nine categories derived from real user requests and exact ground truth from known 3D scene parameters. Because every edit is generated by a known intervention on a controlled scene, EditJudge-Bench evaluates VLM judges as binary verifiers of whether an instruction was correctly executed, isolating judge perception from editor quality. Across 25 judge configurations, we find substantial reliability gaps: (a) the widely adopted VIEScore template reaches only 69.0% AUROC with GPT-4o; (b) fine-tuned judges score 51.7-66.3%; spatial edits such as rotation and shape remain challenging; (c) prompt-formatting choices like image-text ordering shift performance by up to 14.1 points. Despite using only synthetic data, judge rankings on EditJudge-Bench correlate strongly with two real-image human-preference benchmarks ($\rho$=0.95 and $\rho$=0.71), indicating that controlled audits surface failure modes relevant to real evaluation.
"System 1" decision models (pioneered by TypeSafe AI's Jev) replace autoregressive text generation with a single forward pass returning typed probability distributions over declared schemas (Choice, Score, Noul). While deployed for low-latency routing and automated triage, their statistical calibration and post-hoc explainability remain uncharacterized. We introduce an open, deterministic audit benchmark evaluating System 1 decision models across calibration under label-preserving perturbations, cross-question interference, and post-hoc attribution faithfulness. Evaluating five open-weight models (0.8B to 27B) alongside generative System 2 baselines -- both single-token logprob scoring and structured JSON/CoT decoding -- across NLI (e-SNLI), adversarial syntactic shortcuts (HANS), synthetic clinical triage (N=900), prompt injection defense (Tensor Trust, N=250), and an 86-case logic-gate trigger suite yields five findings: (1) provider confidence is an exact closed-form dispersion transform (1 - H(p)/ln K on sub-10B models; normalized margin on Clef-27B) rather than empirical accuracy; (2) bundling multiple questions in one request induces directional cross-question interference (11.7% to 22.4% binary flips on Tev1-0.8B, vs. 0.1% to 0.2% on Tev1-4B); (3) System 1 models with >= 4B parameters are faster and better calibrated under syntactic shifts (ECE_w = 0.030 to 0.087) than logprob (0.144) and JSON (0.378) baselines; (4) for prompt-injection defense (N=250), >= 4B models achieve 0.988 to 0.998 AUROC with up to 95.2% recall and 0.0% false alarms at 48 to 60 ms; and (5) single-pass inference makes 512-coalition XAI interactive (1.3 to 3.2 s/input), while KernelSHAP with a logit link (S1-SHAP) mitigates softmax saturation on disjunctive OR gates (89.3% to 100.0% gate recovery vs. 57.1% to 82.1% in prob space; p_adj = 0.004).
Accurate prediction of electronic Hamiltonians would enable broad property inference while avoiding the high computational cost of Density Functional Theory (DFT). However, progress toward general-purpose materials foundation models is limited by a data bottleneck: existing Hamiltonian datasets are typically small, lack structural diversity, and often omit essential relativistic physics such as spin--orbit coupling (SOC). We therefore construct UniHam, a large-scale Hamiltonian dataset and benchmark suite comprising 100,000+ DFT-computed complex-valued Hermitian Hamiltonians with full SOC, covering 72 elements and a wide range of crystal geometries and symmetries (spanning diverse lattice types and space-group families). Building on UniHam, we benchmark two representative state-of-the-art models under a standardized protocol and introduce complementary evaluation metrics that jointly assess three dimensions: (i) Hamiltonian reconstruction accuracy, (ii) out-of-distribution (OOD) generalization across composition/symmetry shifts, and (iii) the ability to support downstream property prediction from the predicted Hamiltonians. Experiments on UniHam demonstrate that the proposed benchmark and metrics effectively differentiate model capabilities, revealing intrinsic SOC- and element-dependent failure modes, large variations in compositional OOD robustness, and the necessity of spectral-level evaluation to assess whether Hamiltonian predictions reliably support downstream electronic-structure properties. Overall, UniHam provides a reproducible, SOC-complete benchmark that can sharpen model comparisons and accelerate the development of next-generation foundation models for quantum materials.