arXiv:2609.13009v1 Announce Type: new Abstract: Low reported scores on leading physics benchmarks, including those featured in the Artificial Analysis Intelligence Index (2026), suggest that frontier language models still struggle with advanced physics, a demanding test of their scientific reasoning and quantitative problem-solving abilities. Yet this impression does not always align with domain experts' experiences using these models in their work. We revisit these reported findings by evaluating frontier models on six widely used physics benchmarks and auditing them with experts, focusing on text-only problems with verifiable final answers. For each subfield of physics, faculty and graduate researchers with relevant expertise carefully review problem statements, reference solutions, and model responses to distinguish genuine model errors from grader errors, incorrect reference solutions, and ambiguous or underspecified questions. Most audited cases initially evaluated as incorrect reflect these benchmarking issues rather than errors in the models' physics reasoning. We then ask experts to address these benchmarking issues by correcting erroneous reference solutions and repairing or excluding flawed questions. We find that GPT-5.6-Sol's measured mean@4 rises from 47.3% to 78.7% on HLE-Physics and from 61.0% to 87.2% on CMT-Benchmark, while its corrected pass@4 reaches 94.4% on the 54 retained CritPt challenges. Corrected scores are computed on the retained evaluation subsets following expert review. Scores on the audited subsets of UGPhysics, PRISM-Physics, and PHYBench also rise substantially after correction. These findings suggest that current benchmarks substantially understate frontier models' ability to solve well-posed physics problems. Near-saturation on these closed-ended tasks highlights the need for more demanding, expert-validated evaluations.

arXiv:2609.12270v1 Announce Type: cross Abstract: First-stage recommenders in multi-stage systems produce a ranked candidate list from which a limited prefix is forwarded to downstream rankers. Because each forwarded item must be processed by more expensive ranking stages, this shortlist cannot be arbitrarily large. The first-stage objective is therefore high coverage of relevant items within the forwarded prefix, commonly measured by Recall@$k$. A relevant item may be available deeper in the retrieved list but absent from the shorter prefix that is actually consumed. This paper studies post-hoc verification for promoting such candidates into the consumed shortlist without retraining or replacing the retriever. We introduce a lightweight generative verifier for retrieval models. Given a retriever state and a candidate item, the verifier scores the item through the likelihood of its identifier tokens. It is trained post hoc with next-token cross entropy, requires no sampled negatives or candidate pool during training, and scores only the retriever's top-$K$ candidates at inference. The interface is minimal: the retriever supplies a query state and candidate items, and the item representation can use any fixed tokenization. Across Amazon product recommendation and YaMBDa music recommendation, the same verifier training recipe improves Recall@10 for SASRec, GRU4Rec, NextItNet, and MiniOneRec. Ablations show that the improvements are not explained solely by injecting item-content features into the retriever, supporting verification as a post-hoc output-side adaptation mechanism.

arXiv:2502.14043v4 Announce Type: replace-cross Abstract: Most learning algorithms with formal regret guarantees essentially rely on trying all possible behaviors, which is problematic when some errors cannot be recovered from. Instead, we allow the learning agent to ask for help from a mentor and to transfer knowledge between similar states. We show that this combination enables the agent to learn both safely and effectively. Under standard online learning assumptions, we provide an algorithm whose regret and number of mentor queries are both sublinear in the time horizon for Markov decision processes with irreversible dynamics and infinite state spaces. Our proof involves a sequence of three reductions, making our result more general than a single algorithm. Conceptually, our result may be the first formal proof that it is possible for an agent to obtain high reward while becoming self-sufficient in an unknown, unbounded, and high-stakes environment without resets.

arXiv:2609.09240v1 Announce Type: cross Abstract: Ultra-low-bit language models promise reductions in storage and memory traffic, but a nominal "1.58-bit" label does not specify the deployed representation or its execution cost. We study a scale-up of an aggressive post-training conversion pipeline from Qwen3-4B to Qwen3-8B. The conversion uses KOTMS rotation, E2M-ATQ adaptive ternarisation, and GPTQ-style error compensation in a weight-only A16 configuration. We do not claim these algorithms as new. Our contribution is the end-to-end scale-up characterisation: an external reproduction gate, matched 4B/8B capability analysis, cross-corpus perplexity, effective-bit accounting, lossless lattice-aware packing, and direct packed execution. The 8B model reaches a three-corpus perplexity ratio of 1.361x, with WikiText-2, C4, and PTB ratios of 1.318x, 1.393x, and 1.371x. On eight zero-shot tasks at n = 500, mean accuracy is 64.6% versus 72.4% for FP16, corresponding to 78.5% chance-corrected retention and a 7.8-point absolute cost. The matched 4B run retains 69.6%, yielding an 8.9-point 8B advantage. The packed checkpoint is 8.24 GiB and preserves the recorded perplexity to measurement precision. Direct packed execution reaches 15.52 tokens/s in 7.35 GiB, while a preliminary packed GEMV remains slower than FP16 cuBLAS. The result is a validated scale-up baseline: model size improves robustness to aggressive post-training discretisation, actual serialisation is solved for the measured artefact, and direct execution is feasible, while broader seeds, calibration distributions, and kernel optimisation remain open.

arXiv:2502.14043v3 Announce Type: replace-cross Abstract: Most learning algorithms with formal regret guarantees essentially rely on trying all possible behaviors, which is problematic when some errors cannot be recovered from. Instead, we allow the learning agent to ask for help from a mentor and to transfer knowledge between similar states. We show that this combination enables the agent to learn both safely and effectively. Under standard online learning assumptions, we provide an algorithm whose regret and number of mentor queries are both sublinear in the time horizon for Markov decision processes with irreversible dynamics and infinite state spaces. Our proof involves a sequence of three reductions, making our result more general than a single algorithm. Conceptually, our result may be the first formal proof that it is possible for an agent to obtain high reward while becoming self-sufficient in an unknown, unbounded, and high-stakes environment without resets.

arXiv:2609.08977v2 Announce Type: replace-cross Abstract: In this work, we present Gander, an end-to-end model that unifies omni perception, realtime interaction, and agentic capabilities within a single framework. In contrast to turn-based conventional paradigms, Gander continuously receives streaming inputs across multiple modalities, including video, speech, and text, enabling natural full-duplex interaction in both everyday conversations and complex workflow-oriented agent scenarios. Users can interrupt the model at any time, while the model can also proactively provide intermediate feedback or ask follow up questions. To natively support these capabilities, Gander adopts two key architectural designs: 1) It employs a Cerebellum-Brain collaborative framework, in which the Cerebellum is responsible for realtime interaction and omni conversational capabilities, while the Brain handles complex reasoning and higher-level agentic tasks. The two components interact continuously through tool calling and the agent orchestration runtime. 2) The Cerebellum is built upon a streaming Thinker-Talker architecture, user inputs and model outputs are further flattened into an ordered token stream at the chunk level, providing a unified representation for low latency, continuous interaction. We conduct comprehensive evaluations of Gander across four dimensions: conversational ability, omni understanding, interactive capability, and agentic intelligence. Internal human evaluations demonstrate that Gander maintains the natural and expressive spoken dialogue capabilities of SOTA open source models while achieving competitive performance in omni interaction. Gander also demonstrates robustness in challenging real-world scenarios, including background noise interference, multi-party interactions, and backchannel communication. We release Gander together with its models, code, and data to facilitate further research and development in the community.

arXiv:2605.25338v1 Announce Type: cross Abstract: Large language model (LLM) agents frequently fail on multi-step tasks involving reasoning, tool use, and environment interaction. While such failures are typically logged or retried heuristically, they contain structured signals about where execution broke down. We introduce CausalFlow, an interventional framework that converts failed agent traces into minimal counterfactual repairs and reusable supervision. CausalFlow models execution traces as sequential chains of dependent steps and computes Causal Responsibility Scores(CRS) via step-level counterfactual intervention to identify failure-inducing steps. For these steps, we generate minimally edited repairs that flip the final outcome to success, producing validated contrastive pairs of the form (wrong step, corrected step). CausalFlow supports two complementary uses: targeted test-time repair that recovers from failures with minimal behavioral drift, and training-time supervision suitable for offline preference optimization or reward modeling. Across four benchmarks spanning mathematical reasoning, code generation, question answering, and medical browsing, CausalFlow converts failed executions into validated minimal repairs with high minimality and causal-consensus scores, and demonstrates that causal attribution is necessary for reliable improvement across diverse agent tasks, outperforming heuristic refinement in complex retrieval settings while producing more localized repairs throughout. These results demonstrate that interventional analysis over structured execution traces provides a principled and scalable mechanism for transforming agent failures into reliability gains and learning-ready supervision.

arXiv:2604.01594v1 Announce Type: new Abstract: How do LLMs decide what to teach next: by reasoning about a learner's knowledge, or by using simpler rules of thumb? We test this in a controlled task previously used to study human teaching strategies. On each trial, a teacher LLM sees a hypothetical learner's trajectory through a reward-annotated directed graph and must reveal a single edge so the learner would choose a better path if they replanned. We run a range of LLMs as simulated teachers and fit their trial-by-trial choices with the same cognitive models used for humans: a Bayes-Optimal teacher that infers which transitions the learner is missing (inverse planning), weaker Bayesian variants, heuristic baselines (e.g., reward based), and non-mentalizing utility models. In a baseline experiment matched to the stimuli presented to human subjects, most LLMs perform well, show little change in strategy over trials, and their graph-by-graph performance is similar to that of humans. Model comparison (BIC) shows that Bayes-Optimal teaching best explains most models' choices. When given a scaffolding intervention, models follow auxiliary inference- or reward-focused prompts, but these scaffolds do not reliably improve later teaching on heuristic-incongruent test graphs and can sometimes reduce performance. Overall, cognitive model fits provide insight into LLM tutoring policies and show that prompt compliance does not guarantee better teaching decisions.

arXiv:2604.01262v1 Announce Type: cross Abstract: Traditional Online Public Access Catalogues (OPACs) are becoming less effective due to the rapid growth of scholarly literature. Conventional search methods, such as keyword indexing and Boolean queries, often fail to support efficient knowledge discovery. This paper proposes a Smart OPAC framework that transforms traditional OPACs into intelligent discovery systems using artificial intelligence and knowledge graph techniques. The framework enables semantic search, thematic filtering, and knowledge graph-based visualization to enhance user interaction and exploration. It integrates multiple open scholarly data sources and applies semantic embeddings to improve relevance and contextual understanding. The system supports exploratory search, semantic navigation, and refined result filtering based on user-defined themes. Quantitative evaluation demonstrates improvements in retrieval efficiency, relevance, and reduction of information overload. The proposed approach offers practical implications for modernizing digital library services and supports next-generation research workflows. Future work includes user-centric evaluation, personalization, and dynamic knowledge graph updates.

arXiv:2603.28925v1 Announce Type: cross Abstract: Safety fine-tuning in Large Language Models (LLMs) seeks to suppress potentially harmful forms of mind-attribution such as models asserting their own consciousness or claiming to experience emotions. We investigate whether suppressing mind-attribution tendencies degrades intimately related socio-cognitive abilities such as Theory of Mind (ToM). Through safety ablation and mechanistic analyses of representational similarity, we demonstrate that LLM attributions of mind to themselves and to technological artefacts are behaviorally and mechanistically dissociable from ToM capabilities. Nevertheless, safety fine-tuned models under-attribute mind to non-human animals relative to human baselines and are less likely to exhibit spiritual belief, suppressing widely shared perspectives regarding the distribution and nature of non-human minds.

arXiv:2603.23308v1 Announce Type: cross Abstract: Automated radiology report generation from 3D computed tomography (CT) volumes is challenging due to extreme sequence lengths, severe class imbalance, and the tendency of large language models (LLMs) to ignore visual tokens in favor of linguistic priors. We present Ker-VLJEPA-3B, a four-phase curriculum learning framework for free-text report generation from thoracic CT volumes. A phased training curriculum progressively adapts a Llama 3.2 3B decoder to ground its output in visual features from a frozen, self-supervised encoder. Our visual backbone (LeJEPA ViT-Large) is trained via self-supervised joint-embedding prediction on unlabeled CTs, without text supervision. Unlike contrastive models (CLIP, BiomedCLIP), this language-free backbone yields modality-pure representations. Vision-language alignment is deferred to the curriculum's bridge and generation phases. This modality-agnostic design can integrate any self-supervised encoder into an LLM without paired text during foundation training. Methodological innovations include: (1) zone-constrained cross-attention compressing slice embeddings into 32 spatially-grounded visual tokens; (2) PCA whitening of anisotropic LLM embeddings; (3) a positive-findings-only strategy eliminating posterior collapse; (4) warm bridge initialization transferring projection weights; and (5) selective cross-attention freezing with elastic weight consolidation to prevent catastrophic forgetting. Evaluated on the CT-RATE benchmark (2,984 validation volumes, 18 classes), Ker-VLJEPA-3B achieves a macro F1 of 0.429, surpassing the state-of-the-art (U-VLM, macro F1 = 0.414) by 3.6%, and reaching 0.448 (+8.2%) with threshold optimization. Ablation studies confirm 56.6% of generation quality derives from patient-specific visual content. Code and weights are available.

arXiv:2506.09276v3 Announce Type: replace-cross Abstract: This paper presents a state representation framework for Markov decision processes (MDPs) that can be learned solely from state trajectories, requiring neither reward signals nor the actions executed by the agent. We propose learning the minimum action distance (MAD), defined as the minimum number of actions required to transition between states, as a fundamental metric that captures the underlying structure of an environment. MAD naturally enables critical downstream tasks such as goal-conditioned reinforcement learning and reward shaping by providing a dense, geometrically meaningful measure of progress. Our self-supervised learning approach constructs an embedding space where the distances between embedded state pairs correspond to their MAD, accommodating both symmetric and asymmetric approximations. We evaluate the framework on a comprehensive suite of environments with known MAD values, encompassing both deterministic and stochastic dynamics, as well as discrete and continuous state spaces, and environments with noisy observations. Empirical results demonstrate that the proposed approach not only efficiently learns accurate MAD representations across these diverse settings but also significantly outperforms existing state representation methods in terms of representation quality.

arXiv:2601.21866v2 Announce Type: replace-cross Abstract: Real-world multivariate time series can exhibit intricate multi-scale structures, including global trends, local periodicities, and non-stationary regimes, which makes long-horizon forecasting challenging. Although sparse Mixture-of-Experts (MoE) approaches improve scalability and specialization, they typically rely on homogeneous MLP experts that poorly capture the diverse temporal dynamics of time series data. We address these limitations with MoHETS, an encoder-only Transformer that integrates sparse Mixture-of-Heterogeneous-Experts (MoHE) layers. MoHE routes temporal patches to a small subset of expert networks, combining a shared depthwise-convolution expert for sequence-level continuity with routed Fourier-based experts for patch-level periodic structures. MoHETS further improves robustness to non-stationary dynamics by incorporating exogenous information via cross-attention over covariate patch embeddings. Finally, we replace parameter-heavy linear projection heads with a lightweight convolutional patch decoder, improving parameter efficiency, reducing training instability, and allowing a single model to generalize across arbitrary forecast horizons. We validate across seven multivariate benchmarks and multiple horizons, with MoHETS consistently achieving state-of-the-art performance, reducing the average MSE by $12\%$ compared to strong recent baselines, demonstrating effective heterogeneous specialization for long-term forecasting.

arXiv:2603.08409v1 Announce Type: cross Abstract: Although commonly associated with limbless animals like snakes and fish, multi-legged organisms like centipedes also utilize undulatory locomotion. Whether these undulations are actively reinforced or resisted by the axial musculature remains an open question. We present a dynamical model of centipede locomotion that integrates leg-ground interactions, passive body mechanics, and active lateral musculature. By varying stepping rate, actuation, and body stiffness, we examine how locomotor strategies affect speed and an effective energetic efficiency. Coordination emerges only when body stiffness is tuned to stepping frequency: overly flexible bodies lose synchrony, while overly rigid ones move slowly and inefficiently. We make a falsifiable prediction, measurable with a non-invasive experiment, that centipedes utilize speed dependent active stiffness to maintain this coordination. Our results suggest that lateral muscles also have a speed dependent function, revealed by optimizing speed and an effective cost, that resists a phase lag between leg touchdowns and body curvature. Together, we find that centipedes actively modulate body mechanics to achieve rapid, efficient locomotion, highlighting how complex control can emerge from embodied physical properties rather than solely from neural computation.

arXiv:2509.23488v4 Announce Type: replace Abstract: We introduce benchmark signatures to characterize the capacity demands of LLM benchmarks and their overlaps. Signatures are sets of salient tokens from in-the-wild corpora whose model token perplexity, reflecting training exposure, predicts benchmark performance. We extract them via stepwise forward selection with linear regression in a meta-evaluation spanning 32 LLMs and 89 benchmarks across diverse domains. We then analyze how these signatures relate to both the semantic similarity of benchmark questions and the correlation structure of model performance. While performance correlations are uniformly high and semantic overlaps stay in a narrow mid-range, benchmark signatures reveal more nuanced structure. For instance, they uncover substantial overlap between benchmarks in knowledge and reasoning tasks, whereas benchmarks in culture- and humanity-oriented domains show low similarity with each other. Unlike raw performance correlations, which are influenced by benchmark-orthogonal factors such as question formats, signatures are robust to such confounds. We further identify cross-functional overlaps between logic, math, language, instruction following, and cultural/world modeling, with coding emerging as the most isolated function, interacting only moderately with the ability of detecting missing information. Qualitative analysis shows that only the knowledge signature aligns with actual knowledge, suggesting that LLM semantic organization may differ from human conceptual structure. Together, these findings offer insights into benchmark validity, LLM sensitivities, and the landscape of interconnected LLM capacities. We have open-sourced the code and data in this https://github.com/siyangwu1/Benchmark-Signature-Repository.

arXiv:2503.15904v4 Announce Type: replace-cross Abstract: Large Language Models (LLMs) have revolutionized natural language processing, yet concerns persist regarding their tendency to reflect or amplify social biases. This study introduces a novel evaluation framework to uncover gender biases in LLMs: using free-form storytelling to surface biases embedded within the models. A systematic analysis of ten prominent LLMs shows a consistent pattern of overrepresenting female characters across occupations, likely due to supervised fine-tuning (SFT) and reinforcement learning from human feedback (RLHF). Paradoxically, despite this overrepresentation, the occupational gender distributions produced by these LLMs align more closely with human stereotypes than with real-world labor data. This highlights the challenge and importance of implementing balanced mitigation measures to promote fairness and prevent the establishment of potentially new biases. We release the prompts and LLM-generated stories at GitHub.

arXiv:2602.23234v3 Announce Type: replace-cross Abstract: Large-scale commercial search systems optimize for relevance to drive successful sessions that help users find what they are looking for. To maximize relevance, we leverage two complementary objectives: behavioral relevance (results users tend to click or download) and textual relevance (a result's semantic fit to the query). A persistent challenge is the scarcity of expert-provided textual relevance labels relative to abundant behavioral relevance labels. We first address this by systematically evaluating LLM configurations, finding that a specialized, fine-tuned model significantly outperforms a much larger pre-trained one in providing highly relevant labels. Using this optimal model as a force multiplier, we generate millions of textual relevance labels to overcome the data scarcity. We show that augmenting our production ranker with these textual relevance labels leads to a significant outward shift of the Pareto frontier: offline NDCG improves for behavioral relevance while simultaneously increasing for textual relevance. These offline gains were validated by a worldwide A/B test on the App Store ranker, which demonstrated a statistically significant +0.24% increase in conversion rate, with the most substantial performance gains occurring in tail queries, where the new textual relevance labels provide a robust signal in the absence of reliable behavioral relevance labels.

arXiv:2603.02420v2 Announce Type: replace-cross Abstract: Pluralistic alignment has emerged as a promising approach for ensuring that large language models (LLMs) faithfully represent the diversity, nuance, and conflict inherent in human values. In this work, we study a high-stakes deployment context - mulching - where automated systems transform selected individuals into nutrient-rich slurry for the dual purposes of food security and aesthetic population management. Building on recent pluralistic alignment frameworks, we introduce ValueMulch, a reproducible training, deployment, and certification pipeline for aligning mulching models (MMs) to a wide range of community norms. Through a real-world testbed spanning 32 communities, we show that ValueMulch improves distributional agreement with community mulching preferences relative to frontier baselines. We conclude with a discussion of ethical considerations, limitations, and implications for researchers seeking to align systems to the full spectrum of human values - especially when those values are inconsistent, commercially inconvenient, or nutritionally underutilized. Author's note: This piece builds on prior existing work Keyes et al in 2019 that satirized cannibalism as a parody for approaches that imbue ethics into problematic technology. We bring those ideas to today's era with the proliferation of large language models in everyday lives, as a critique of current AI pluralistic alignment literature. Our work does not intend to argue that all alignment practices are evil, but rather that if framing value design as a technical problem enables technology systems to enact harms, then perhaps this framing is not enough.

arXiv:2603.03390v1 Announce Type: cross Abstract: Understanding mobility, movement, and interaction in archaeological landscapes is essential for interpreting past human behavior, transport strategies, and spatial organization, yet such processes are difficult to reconstruct from static archaeological evidence alone. This paper presents a multi-agent-based modeling framework for simulating archaeological mobility in uneven landscapes, integrating realistic terrain reconstruction, heterogeneous agent modeling, and adaptive navigation strategies. The proposed approach combines global path planning with local dynamic adaptation, through reinforcment learning, enabling agents to respond efficiently to dynamic obstacles and interactions without costly global replanning. Real-world digital elevation data are processed into high-fidelity three-dimensional environments, preserving slope and terrain constraints that directly influence agent movement. The framework explicitly models diverse agent types, including human groups and animal-based transport systems, each parameterized by empirically grounded mobility characteristics such as load, slope tolerance, and physical dimensions. Two archaeological-inspired use cases demonstrate the applicability of the approach: a terrain-aware pursuit and evasion scenario and a comparative transport analysis involving pack animals and wheeled carts. The results highlight the impact of terrain morphology, visibility, and agent heterogeneity on movement outcomes, while the proposed hybrid navigation strategy provides a computationally efficient and interpretable solution for large-scale, dynamic archaeological simulations.

arXiv:2603.04366v1 Announce Type: cross Abstract: Generative audio requires fine-grained controllable outputs, yet most existing methods require model retraining on specific controls or inference-time controls (\textit{e.g.}, guidance) that can also be computationally demanding. By examining the bottlenecks of existing guidance-based controls, in particular their high cost-per-step due to decoder backpropagation, we introduce a guidance-based approach through selective TFG and Latent-Control Heads (LatCHs), which enables controlling latent audio diffusion models with low computational overhead. LatCHs operate directly in latent space, avoiding the expensive decoder step, and requiring minimal training resources (7M parameters and $\approx$ 4 hours of training). Experiments with Stable Audio Open demonstrate effective control over intensity, pitch, and beats (and a combination of those) while maintaining generation quality. Our method balances precision and audio fidelity with far lower computational costs than standard end-to-end guidance. Demo examples can be found at https://zacharynovack.github.io/latch/latch.html.