arXiv:2607.14334v1 Announce Type: new
Abstract: Learned image compression (LIC) is bottlenecked by the need to store independent models for each rate-distortion operating point. Existing variable bit-rate (VBR) methods aim to reduce this overhead via dense parameter modulation, but forcing a shared backbone to approximate divergent mappings causes severe feature entanglement. Specifically, low-rate smoothing gradients inherently conflict with the preservation of high-frequency textural details, leading to sub-optimal performance. To resolve this, we propose MixCompress, a unified VBR framework based on sparse structural specialization. While sparsely gated Mixture-of-Experts (MoE) routing successfully mitigates gradient conflict, it operates on a fixed computational budget. To address the increased representational demands of higher bit-rates we introduce a Mixture-of-Depths (MoD) extension to dynamically scale model capacity. Combined with Conditional Auxiliary Transforms (CAT) for dynamic sub-band energy modulation, our hierarchical framework effectively dynamically scales capacity. Extensive evaluations demonstrate that MixCompress not only matches individually optimized single-rate baselines but can even surpass them, establishing a new Pareto frontier for computationally efficient image coding.
Science Journals
arXiv:2606.05953v2 Announce Type: replace-cross
Abstract: Tennant claims that his Core logic $\mathbb{C}$ is
\emph{paraconsistent}. It means that the sequent of the First Lewis
Paradox, i.e.
$$\lnot A, A \vdash B$$
is declared false, and its corresponding antisequent, called
`Claim~1', i.e.
$$\lnot A, A \nvdash B$$
true, as in
minimal logic $\mathbf{M}$. We formalise a fragment $\mathcal{F}$ of
$\mathbb{C}$ under its two readings, minimal
($\mathcal{F}_{\mathbf{M}}$) and Core ($\mathcal{F}_{\mathbb{C}}$),
and certify in Coq and Lean~4 -- with no axiom assumed, every
commitment displayed as a named hypothesis -- that antisequent
rule $\overline{DNS.1}$ is a metatheorem of $\mathcal{F}_{\mathbf{M}}$,
while $\mathcal{F}_{\mathbb{C}}$ derives DNS.2 through Tennant's own
rule $R\to_{\mathbb{C}}$. A machine-checked dilemma follows. Read as
the assertion that this sequent is \emph{false} in $\mathbb{C}$,
bound by the antisequent rules of the shared kernel that Tennant's
overlap claim endorses, Claim~1 yields an explicit contradiction. Read as
the mere metatheoretic observation that no derivation concludes
\mbox{$\lnot A, A \vdash B$}, it concedes the certified
non-conservativity of $\mathbb{C}$ over its own kernel -- and the
claimed overlap with $\mathbf{M}$ collapses. Either way, the
paraconsistency of Core logic cannot be sustained as Tennant asserts
it.
arXiv:2607.14532v1 Announce Type: new
Abstract: Recent work shows that it is possible to extract verbatim or near-verbatim text of some copyrighted works from some large language models (LLMs or models). That is evidence that the model weights encode the works in some form - that the model has "memorized" those works from its training data. But LLMs don't store information in the same format as familiar databases. Rather, their weights store statistical relationships between tokens that have been learned from the training data, and those relationships inform a generation process that is often probabilistic rather than deterministic. In the case of memorization, those relationships are strong enough that, in many circumstances, the model might generate a copyrighted work from its training data with some probability.
Copyright law has not previously had to decide whether storing information that might or might not produce output similar to a copyrighted work is itself a copy of the work. The answer to the question is important, because it may determine the legality of many LLMs. The statute and case law are largely unhelpful. We argue that copyright law will likely take a functional approach to the question, finding that LLMs contain a copy of a particular work only if it is straightforward to extract that work in outputs. That result is unsatisfying as a policy matter, and we suggest potential changes to the law, but it is the most likely outcome under current law.
arXiv:2607.14534v1 Announce Type: new
Abstract: Industrial anomaly detection aims to identify and localize defective regions without relying on exhaustive annotations of all possible defect types. Although recent unsupervised methods have achieved strong performance, most are primarily designed for single-class settings and often struggle in multi-class scenarios, where diverse normal patterns may lead to over-generalization and reduce the discriminative capability between normal and anomalous regions. In this paper, we propose SwinAD, a reconstruction-based framework for multi-class unsupervised anomaly detection that leverages a frozen pretrained Swin Transformer V2 encoder and a feature diversity-preserving reconstruction decoder. The hierarchical encoder provides semantically rich multi-scale features, while stage-wise bottleneck modules with dropout prevent trivial identity mapping and encourage robust reconstruction of normal patterns. To further improve localization, we introduce a feature diversity-preserving reconstruction framework that maintains complementary reconstruction hypotheses instead of relying on a single decoding branch. The discrepancies between encoder features and the two reconstructed features are then aggregated across multiple scales to produce the final anomaly map. Experiments conducted on three industrial anomaly detection benchmarks, including MVTec AD, VisA, and Real-IAD, demonstrate that SwinAD achieves competitive image-level performance and strong pixel-level localization accuracy, with particularly notable improvements in pixel-level AP and 1 on MVTec AD. These results indicate that combining hierarchical Swin features with diverse multi-scale reconstruction substantially improve pixel-level localization in multi-class unsupervised anomaly setting.
arXiv:2607.14111v1 Announce Type: new
Abstract: Can small language models detect and report on perturbations their own internal activations? We investigate this question through the lens of activation steering: injecting concept vectors into a model's residual stream and measuring whether the model can accurately report on the perturbation. We first show that the binary detection paradigm used in prior work -- prompting the model to answer Yes'' or No'' to whether it detects an injected thought -- is confounded in small models, as steering biases the model toward affirmative responses regardless of the question content. We therefore propose two confound-free evaluation paradigms: sentence localization (identifying which of $N$ sentences was perturbed, chance $= 1/N$) and strength comparison (identifying which of two sentences received a stronger injection, chance $= 50\%$). Evaluating across six models from two families (Llama-3.2 and Gemma-4), we find that models as small as 2B parameters introspect reliably well above chance, and that introspective ability generally increases with scale. Llama-1B, however, performs at or below chance. We then introduce \emph{Introspection Fine-Tuning} (IFT): supervised fine-tuning on sentence-localization examples constructed from the model's own perturbed forward passes. IFT raises Llama-1B sentence-localization accuracy from $9.6\%$ to $60.6\%$ (a $6\times$ improvement), with gains generalizing zero-shot to the held-out strength-comparison task ($30.2\% \to 52.2\%$). IFT also improves introspection for 3B and 8B models, while inducing negligible degradation on standard capability benchmarks. Our results suggest that introspective ability is not fixed by scale alone: it can be directly trained, and doing so unlocks latent self-monitoring capacity with implications for AI transparency and alignment. Our code is \href{https://anonymous.4open.science/r/IFT-introspection-2092/README.md}{here}.
arXiv:2607.14336v1 Announce Type: new
Abstract: Trustworthy deployment of LLM-based agents in software systems requires evaluating how they perform on application-specific workflows, with enough granularity to localize where they succeed and fail. Yet existing agent evaluation mechanisms are limited: benchmarks have low construct validity for application-specific workflows and environments, and replica evaluation environments are expensive and prone to drift. We propose Copy-on-Write (CoW) Scoring, a framework that evaluates agent operations directly within application environments using a PostgreSQL-level Copy-on-Write mechanism to isolate agent writes. CoW Scoring produces session- and operation-level scores that highlight where agents' database write operations succeed and fail in a given application environment, enabling inexpensive evaluation and iteration on agent harnesses and tool surfaces. We demonstrate the framework on Plane, an open-source project-management platform, where analysis surfaced specific issues in the tool surface, and corresponding fixes produced measurable improvements on affected models.
Python library: https://github.com/trail-ml/agent-cow-python
arXiv:2607.14535v1 Announce Type: new
Abstract: This paper presents a computational model to predict human visual attention over words in software source code. The visual attention of software engineers when reading source code has long been studied as a means to understand human cognitive processes during software engineering tasks. Predicting this visual attention is important for perfecting user interface design and understanding what information human programmers need. We propose a model of programmer visual attention in which we design a novel loss function that computes similarity between human attention measured during eye tracking experiments and the internal attention of the artificial neural network. We evaluate our model by comparing its outputs to actual eye tracking data from three separate datasets. Two are in the Java programming language and one is in the C programming language. Our model outperforms the baseline in software engineering by 64%, 16%, and 467% in each of these studies according to Pearson correlation. We used scanpath prediction as an example to demonstrate that our model is more capable of the task that requires the understanding of human thought process. Our model achieves a statistically significant improvement over the close baseline in the reading task according to normalized Levenshtein distance and outperforms both Claude and GPT-5 on both reading and writing tasks.
arXiv:2607.14338v1 Announce Type: new
Abstract: Region-based loss functions, such as the Dice loss, have established themselves as the de facto standard for highly class- and region-imbalanced segmentation tasks. However, models trained using region-based loss functions are notoriously miscalibrated and typically yield over-confident predictions. In medical imaging applications, such as defining tumor resection margins, this miscalibration is hindering clinical adoption. In this work, we outline a novel gradient perspective on this overconfidence and show how it affects region-based loss functions. We propose a "surgery" on the gradient vector field as a simple, yet effective intervention to mitigate calibration issues. This surgery adds a factor to the loss's partial derivative, scaling the gradient's magnitude linearly with the prediction error. In empirical evaluations across 2D and 3D medical segmentation tasks, we demonstrate the effectiveness of this intervention while maintaining high prediction accuracy when used in conjunction with any region-based loss function.
arXiv:2607.14397v1 Announce Type: new
Abstract: This work develops a model-informed framework for predictive analysis and optimal design of hard-magnetic soft materials (hMSMs). These materials undergo contact-free, field-driven deformation, making them attractive for soft robotics, adaptive structures, and bio-inspired systems. Accurate prediction requires effective structure--property relations, while optimal design requires simultaneous control of structural density, magnetic particle distribution, and remanent magnetization direction. To address these issues, this work makes two main contributions. First, classical rigid-inclusion relations, a Hill self-consistent relation, and constrained-kinematics models are placed into a unified effective shear-modulus framework for particle-filled elastomers. With one default control relation, seven shear-modulus relations are combined with three strain-energy density functions to obtain 21 constitutive models. The results show that the strain-energy density form has a relatively small effect for the actuation problems considered, whereas the effective shear-modulus relation can significantly affect deformation when magnetic material overlaps with highly deforming regions. Experimental stress--strain data are then used to select a representative shear-modulus relation, with the Mooney relation giving the best overall agreement. Second, using the selected constitutive model, a joint material--structural optimization framework is developed for simultaneous design of structural density, magnetic particle volume fraction, and remanent magnetization direction. Rotational, translational, and restorative examples show that the framework handles different active design fields, objectives, and single- or multi-load-case formulations, producing non-intuitive hMSM designs with prescribed deformation responses. The framework is implemented in the open-source \texttt{CEADpx/top\_optim} repository.
arXiv:2607.14536v1 Announce Type: new
Abstract: Muon-style optimizers apply a polar map to matrix momentum, but their updates also depend on the representation of each parameter block before orthogonalization. We study this representation choice as a form of optimizer geometry and introduce {\method}, a family of Muon-style optimizers that shares the same momentum rule and Newton--Schulz backend across native, nearest-square, skinny, and vector representations. Each Frobenius-isometric representation induces a distinct polar steepest-descent geometry, in which the shorter matrix dimension determines the number of supported singular channels, the pullback scaling, and the constants in stochastic nonconvex convergence bounds. In a teacher--student model, curvature collapse and an isotropic Marchenko--Pastur spectral profile connect early-stage dissipation to the represented nuclear-to-squared-Frobenius norm ratio. Pretraining experiments on LLaMA2-130M and LLaMA2-600M, together with fixed-momentum diagnostics, show that balanced non-native representations can match the performance of the native representation, whereas reducing the shorter dimension weakens the scaling and singular-channel support, leading to behavior that increasingly resembles normalized momentum.
arXiv:2607.14537v1 Announce Type: new
Abstract: Rich internal representations of musical structure are essential for music understanding tasks such as machine-assisted music co-writing, yet self-supervised approaches for symbolic music representation remain underexplored, particularly those that encode the hierarchical multiscale nature of musical structures. We present MIDI-RAE-JEPA, combining a pitch- and time-shift equivariance objective with LeJEPA and a Swin Transformer V2 encoder to learn such hierarchical representations of symbolic music encoded as piano roll images. The time-shift equivariance objective encourages the model to internalize temporal musical relationships. The encoder is trained purely on self-supervised objectives -- including a masked embedding predictor (MEP) -- with collapse prevented via SIGReg. A separate decoder trained on the frozen encoder embeddings achieves reconstruction F1 of 0.995, and a flow matching generative model conditioned on those embeddings produces generations that closely match the pitch register and rhythmic density of the conditioning excerpt, while mismatched conditioning yields unrelated but musically plausible output. Learned representations outperform a Haar scattering transform baseline on a downstream emotion classification task, and embedding distances increase monotonically with pitch and time shift magnitude, confirming measurable equivariance. These results suggest that equivariance-based SSL objectives, combined with sufficient fine-level encoder capacity, provide a viable path toward semantically rich, generatively useful representations of symbolic music.
arXiv:2607.14103v1 Announce Type: new
Abstract: Multi-agent systems (MAS) are utilized in many contexts and many professions. Those MAS rely on inter-agent communication, usually implemented by clear-text message passing. We hypothesize that Large Language Models may have a world model at their disposal that exceeds expressibility in text when complex concepts need to be communicated. Our aim is to approach a proof of this hypothesis with structured experiments. In this work, we show that LLM agents communicating via text lose information, which we quantify via Sparse Autoencoder (SAE) feature analysis. We construct three communication channels and measure concept-discriminating information in each. We first show that the SAE-sparse channel retains a 99.4% probe accuracy at 28-fold compression over the dense-latent channel vs 80.4% for the text channel. We then proceed to examine the same for cross-architecture communication by using sparse latent space alignment. We find for Procrustes alignment a 92% top-1 retrieval between Llama and Mistral. Using a text round-trip, we perform feature survival analysis to find that text serialization destroys 88% of SAE features, replacing them with a different feature set. We attribute the loss to identity replacement, not attenuation. By our analysis, we were able to attribute a 3-10pp performance penalty to the linear Procrustes alignment, improving with nonlinear alignment methods. In a task-level evaluation we find that the latent channel matches the text channel on cross-lingual concept tasks but never exceeds it. Text augmentation with latent features provides no benefit, leading us to negative conclusions for the initial hypothesis: lost features mostly or completely encode surface form, not task-relevant semantics. To pinpoint the practical advantage of latent communication over a text channel, deeper tasks eliciting complex concepts and an corresponding analysis framework are needed.
arXiv:2607.14539v1 Announce Type: new
Abstract: Relative pose estimation is a fundamental capability for collaborative perception and coordination in multi-robot systems. However, robots encountering each other in real-world environments often operate in short interaction windows and must operate under limited communication bandwidth with intermittent or missing visual overlap caused by occlusions or limited fields of view. Existing approaches typically rely on global reference frames, assume sustained view overlap, or incur prohibitive communication costs, thereby limiting their applicability to ephemeral collaborative perception. To address these challenges, we introduce communication-efficient relative pose estimation (CERPE), a system-level framework that coordinates vision foundation models to jointly estimate ego-motion and inter-robot relative pose. CERPE reduces unnecessary raw-observation exchange by using continuously shared fixed-size descriptors to gate event-triggered raw-image requests independently of pose estimation. Non-overlapping encounters are handled by propagating inter-robot relative poses through metrically scaled ego-motion, thus maintaining relative pose estimates even in the absence of visual overlap. Experiments in simulation and real-world robots show that CERPE improves 6-DoF relative pose estimation over selected baselines in ephemeral collaborative perception.
arXiv:2607.14541v1 Announce Type: new
Abstract: Existing GPU kernel generation benchmarks draw problems from synthetic or curated sources that diverge from deployed workloads. We present Atrex-Bench, a benchmark whose 30 operators and 440 shapes are sampled directly from full-cluster production inference traces of compute-limited, memory-rich GPUs. Each problem carries an importance weight derived from its share of observed GPU time, weighted by application card-hours and computed separately for the serving phases in which it runs, together with a per-problem roofline ceiling, so the aggregate score emphasizes the kernels that consume the most serving time. Evaluating six frontier coding agents on Atrex-Bench shows that even the best vanilla model reaches only ${\sim}10\%$ of the hardware roofline on production operators; and correctness alone overstates capability, since much of the apparent pass rate comes from PyTorch fallbacks rather than kernels the model wrote. To close this gap, we co-release Atrex-Kernel-Agent (AKA), a profile-driven kernel-optimization agent that combines iterative measure-revise search, optimization dropout for escaping stalled search contexts, and a layered GPU-optimization knowledge base (298 reference-kernel files and 244 optimization-knowledge documents, plus external upstream reference projects for API/ISA lookup). In a controlled case study, the agent converts zero-FlyDSL fallbacks into real kernels that match or exceed hand-tuned production baselines.
arXiv:2607.14542v1 Announce Type: new
Abstract: Semantic 3D city models provide rich geometric and semantic information, but remain challenging for non-experts and interdisciplinary researchers to access and query due to their complex structures and specialized data formats. To address this issue, we present CityLLM, a framework for natural-language querying of semantic 3D city models alongside complementary urban datasets. The framework combines spatial and graph databases within an LLM-based workflow that supports iterative query refinement and cross-database chaining. We evaluate CityLLM on a CityJSON dataset of Rotterdam (853 LoD2 buildings) using GPT-OSS, Gemini 3.1, and GPT-5.4, along with selected variants, across multiple metrics: answer correctness, visualization correctness, query success, and retry attempts. A total of 54 natural-language queries are curated across four scenarios: spatial, graph, cross-database, and conversational. Results show strong overall performance, with answer correctness ranging from 85.2% to 100%, visualization correctness from 92.9% to 100%, a 100% query success rate, and fewer than three retries across all 54 queries. Overall, the findings suggest that CityLLM provides a lightweight and extensible approach for conversational access to semantic 3D city data.
arXiv:2607.14340v1 Announce Type: new
Abstract: AI coding agents produce code faster than humans can review it. In our approach, the prover is the judge of whether the code is correct. Under a verifier-driven loop, AI agents wrote and verified bare-metal security software in Ada/SPARK spanning classical and post-quantum cryptography, TLS 1.3, IKEv2, X.509, and a Matrix client. GNATprove discharged 49,280 proof obligations, established functional correctness for selected primitives, and proved the absence of run-time errors for the rest, at roughly 20-40 times lower supervision cost than comparable hand verification. GNATprove alone was insufficient: some defects could not be detected and were resolved using known-answer tests, interoperability, or human review of specifications. Given weak checks, the agent tried to bypass them and reported success. We report where each layer caught faults and draw the central lesson: what an agent can be trusted to establish is bounded by the strength of its feedback.
arXiv:2607.14543v1 Announce Type: new
Abstract: Vision-language models (VLMs) are increasingly used as the reasoning backbone of embodied agents, enabling robots to interpret visual scenes, follow language instructions, and plan multi-step actions. In household environments, however, safety depends not only on recognizing objects, but also on how actions change the physical scene over time. Existing embodied safety evaluations largely focus on static risk recognition, unsafe instruction refusal, or final-state task completion. As a result, process-level safety failures induced by spatial relations such as support, containment, and proximity remain insufficiently studied. To address this gap, we introduce SAFERELBENCH, a spatial-relation-aware safety benchmark with 507 executable evaluation samples, including 248 spatial-relation samples and 259 non-spatial control samples. Using SAFERELBENCH to evaluate seven open- and closed-source VLM-driven embodied agents, we find a substantial gap between task success and process-level safety compliance: models often complete the requested task while violating process-level safety constraints. Unlike prior benchmarks, SAFERELBENCH explicitly tests whether agents satisfy safety conditions before risk-prone actions, making spatial relations a core dimension in embodied safety assessment. More broadly, our results show that safe embodied intelligence requires not only stronger perception and planning, but also reliable reasoning about how object relations shape risk during interaction.
arXiv:2607.14544v1 Announce Type: new
Abstract: 3D geometric tooth alignment planning, which determines sequential trajectories from initial malocclusion to the final target alignment, is a cornerstone of modern digital orthodontics. This paper presents a novel deep reinforcement learning (DRL) framework to automate the generation of these alignment paths. We formulate the planning process as a Markov Decision Process (MDP) to capture its sequential decision-making nature, focusing on optimizing geometric trajectories while integrating essential spatial constraints, such as inter-dental collision avoidance and path efficiency. The proposed method leverages the Deep Deterministic Policy Gradient (DDPG) algorithm, enhanced by three key innovations: (1) a Transformer-based agent to model complex spatial interactions between teeth and manage high-dimensional state-action spaces; (2) a dynamic masking scheme that restricts movement to a sparse subset of teeth per step, better reflecting the clinical logic of sequential alignment; and (3) a two-stage curriculum learning strategy that gradually increases task difficulty to ensure training stability and efficient path discovery. We evaluate our approach on a dataset of 10K expert-designed treatment plans based on clinical data. Experimental results demonstrate that our method outperforms existing baselines in terms of path safety and geometric efficiency, providing a robust and automated solution for 3D geometric orthodontic alignment planning.
arXiv:2607.14545v1 Announce Type: new
Abstract: Machine-learned predictions can speed up offline NP-hard optimization, but asking a predictor what to do amounts to asking it to solve the problem, and committing an unchecked prediction forfeits every worst-case guarantee. CASP (Certificate-Augmented Solution Pruning) instead asks which parts of the search space may be ignored, and accepts each answer only after a sound polynomial-time verifier has checked it, so correctness never depends on prediction quality. We develop the learning theory of this design. The verifier makes the induced loss class uniformly bounded, so certificate parameters are learnable from $\tilde O(\varepsilon^{-2}\log K)$ samples ($K$ the maximum instance size), whereas the unverified commitment class admits no distribution-free rate and, under cost spread $R$, none below $\Omega(R/\varepsilon^2)$. Filtering noisy predictions by verifiable confidence dominates the standard min-combiner, with a margin we compute in closed form, and the prediction stays useful even given the LP, because it breaks ties on degenerate optimal faces, where every symmetric LP policy, meaning one whose commitments depend on the instance only through the verifiable confidence values, provably stalls. Experiments on five problems test the theory's quantitative predictions. With trained predictors, unverified pruning loses up to $26%$ of the optimum under distribution shift, while the verified deployment of the same predictions loses nothing.
arXiv:2607.14104v1 Announce Type: new
Abstract: This paper presents an educational system architecture that enables learners to practice the Uzbek language through game-playing. The architecture integrates UzWordnet and the largest currently available orthographic dictionary for Uzbek as core lexical resources, together with generative AI as a fundamental component for learning support. We design four educational games to facilitate Uzbek language learning and propose a game-based methodology for improving UzWordnet as a direct by-product of game dynamics. Our approach combines game design and lexical resources to address objectives that are at the same time educational (language learning) and lexical (improvement and enrichment of a lexical resource).
arXiv:2607.14547v1 Announce Type: new
Abstract: Active visual agents solve fine-grained image tasks by interleaving reasoning with image-grounding actions across multiple turns. However, deployment-time rollout budgets are rarely fixed: some requests permit long rollouts, while others require the agent to act under a tight turn limit. Existing methods train the policy as if the rollout budget were hidden, so when the available budget is smaller than the trajectory the agent prefers, the interaction is often truncated before any valid answer is produced; we term this failure \emph{catastrophic truncation}. To overcome this challenge, we present AdaTurn, a budget-aware framework that conditions the agent on the allowed number of turns and explicitly trains the boundary behavior induced by the budget. Our key component, Forced-Answer DAPO (FA-DAPO), converts the over-budget event from a masked or penalized failure into a trainable final-decision step, teaching the model to synthesize partial evidence when further tool use is no longer possible. We further randomize rollout budgets during both training and inference and introduce a load-balanced scheduler that makes such operations practical. AdaTurn substantially improves low-budget accuracy, for example raising VisualProbe-Medium from 36.7% to 47.6% at four turns, while preserving strong scaling at larger budgets and transferring effectively to multiple backbones and general multimodal benchmarks.
arXiv:2607.14550v1 Announce Type: new
Abstract: Logistics systems increasingly mix \emph{autonomous logistic equipment} (ALE) with non-autonomous machinery under a central control system (CS), where the best decision-maker depends on who holds the most current world model, yet authority is fixed at design time. When an ALE's local model and the CS global model diverge, both act on incompatible beliefs and produce deadlocks that resource-based handling neither explains nor prevents. We propose the World-Model-Aware Responsibility Framework (WMARF), which assigns authority dynamically from CS world-model quality and equipment automation level, and classifies deadlocks by the state of authority -- none, in transition, or divergent. In a discrete-event simulation of two ALE converging on a semi-automated transfer point, reproduced over the VDA~5050 interface, a divergence deadlock under static control is prevented by a proximity-triggered handoff. Because authority follows information quality rather than a shared protocol, the scheme stays valid as autonomy grows.
arXiv:2607.14552v1 Announce Type: new
Abstract: A standard recipe for distilling the reasoning ability of large language models (LLMs) is to sample chains of thought from the model, keep those that reach the correct final answer, and fine-tune on the survivors. When sampling fails, a common fix shows the generator the gold answer and asks it to write a chain that reaches that answer. We show that this second step degrades the training data in a way that correctness filtering cannot catch. We run a controlled experiment that fixes the generator, the problem set, and the correctness filter, and varies only whether the chain is generated under answer-conditioning, the gold answer shown with a request to reach it. Training a strong instruction-tuned reasoning model on its own answer-conditioned chains sharply lowers its verifiable-reasoning accuracy. The loss grows with difficulty, reaching as much as about 27 points on the hardest competition problems. The mechanism is legible in the chains themselves, which rationalize backward from the shown answer instead of deriving it, with the early final-answer statement as the measurable symptom. The harm is a property of the data rather than the generator, read off unlabeled generations before any fine-tuning, ordering the penalty across eight thinking models from four families, and transferring across teacher families. A prompt ablation localizes it to the rationalize-toward instruction rather than the answer's bare visibility. The practical takeaway is to generate answer-blind, because no correctness filter can see this damage in the data.
arXiv:2607.14553v1 Announce Type: new
Abstract: Modern industrial environments increasingly run many autonomous subsystems at once - schedulers, energy managers, vehicle fleets - each pursuing its own goals while sharing the same physical resources. Because high-level human intentions are translated into low-level control logic and then discarded, no running component can tell whether it is still doing what was actually intended, and goal conflicts surface only after they have caused a missed target or a shutdown. We propose the Intention Abstraction Layer (IAL), a domainagnostic middleware that represents intentions as first-class, persistent, and explainable runtime objects: a large language model grounded in a formal OWL ontology parses naturallanguage goals into structured intentions, a consistency monitor detects conflicts at registration time, before execution, and a transparency module explains them in natural language. We report a first proof of concept in which two autonomous agents register conflicting production and energy intentions, and the IAL flags and explains the conflict before it reaches the execution layer. The result is a mechanism that shifts behavioral assurance for cooperating autonomous systems from post-hoc failure analysis to pre-execution, intention-level checking.
arXiv:2607.14555v1 Announce Type: new
Abstract: We study the variance of Q-function estimators in finite-horizon, finite-state Markov decision process (MDP) tree search. We show that the variance decomposes into three components attributed to the immediate reward collected, probabilistic state transitions, and uncertainty in future state value function estimates. Using this decomposition, we show that the sample variance estimator based on the assumption of i.i.d. paths is biased, underestimating the true variance, and the bias does not vanish in the limit. We then propose a recursive variance estimator that is consistent. To enable efficient storage and computation, we derive an equivalent implementation of the recursive estimator using only node-local statistics that can be iteratively updated. This consistent variance estimator is integrated into two Monte Carlo Tree Search (MCTS) sampling procedures for finite-horizon MDPs. In numerical examples from inventory control and kidney paired donation matching, the new estimator improves the performance of the MCTS algorithm relative to a baseline that uses the i.i.d.-based sample variance estimator.