Academic Intelligence · Curated Daily

探索全球前沿学术脉络

AcademicHub 汇聚顶级期刊与预印本平台的实时文献。定制您的专属科研雷达,利用大语言模型自动生成交叉领域文献分析简报。

01.
Nature (Science) 2026-06-24

Ductile alloys offering 100 MPa tensile strength at 2,400 °C

作者:

Extreme applications call for materials that are not only strong to withstand thermomechanical loads at temperatures in excess of 2,000 °C (refs. 1–3), but also highly formable at room temperature to allow for processing into complex-shaped parts. The latter excludes brittle ceramics4 and intermetallic compounds5, limiting the selection to highly ductile metals and their alloys, but for them, an adequate strength at ultrahigh temperatures seems unreachable. Here we show a breakthrough in casting alloys that achieve both simultaneously. A boron-stabilized HfO2-strengthened Ta-based alloy was carefully crafted using a new boron-intervened in situ oxidation reaction, producing about 50-nm diameter oxide particles dispersed densely and uniformly in the grain interior. The new alloy fills the blank at ultrahigh temperatures in terms of tensile yield strength, around 200 MPa at 2,000 °C and 100 MPa at 2,400 °C, while simultaneously possessing an excellent strength–ductility balance at room temperature (ultimate tensile strength >800 MPa, elongation-to-failure of about 35%), a property combination surpassing all previous refractory (including multi-principal-element) alloys. Moreover, the boron segregation around the oxide nanoparticles imparts excellent thermal stability against coarsening at 2,000–2,400 °C. Our strategy thus goes beyond traditional oxide-dispersion strengthening to enable highly ductile refractory alloys that are capable of load-bearing applications at extreme temperatures. A boron-stabilized oxide-strengthened tantalum alloy combines exceptional room-temperature ductility with record ultrahigh-temperature strength, enabling load-bearing applications above 2,000 °C.

02.
arXiv (CS.AI) 2026-06-18

Maturing Markov Decision Processes: Decision Making under Increasing Information and Shrinking Action Sets

arXiv:2606.18820v1 Announce Type: cross Abstract: Sequential decision problems often exhibit an asymmetric evolution of information and decision flexibility: as a decision cycle unfolds, the agent receives richer information while feasible actions expire due to operational cutoffs, commitments, or resource constraints. Standard MDP formulations typically flatten this structure into stage-dependent state descriptions and action masks, thereby obscuring the nested information–action asymmetry that determines which decisions are urgent and which can be deferred. We introduce Maturing Markov Decision Processes (MMDPs), a formulation built around this information–action asymmetry. We characterize one of its key consequences through an expiring-action priority principle, which identifies the actions that must be resolved before the next stage. Motivated by this structure, we develop a structure-aware reinforcement learning framework with stage-aware policy design, expiring-action abstraction, and search-augmented learning with distillation. Experiments on a controlled multi-supplier replenishment problem, simplified cash-management environments of increasing complexity, and a production-scale simulator show that explicitly modeling this asymmetry improves learning efficiency and becomes increasingly valuable as decision problems scale.

03.
arXiv (CS.CL) 2026-06-18

Graph-ESBMC-PLC: Formal Verification of Graphical PLCopen XML Ladder Diagram Programs Using SMT-Based Model Checking

PLCopen XML defines two encoding formats for IEC 61131-3 Ladder Diagram programs: a textual encoding using elements, and a graphical encoding that represents rung logic as a directed graph of localId/refLocalId connections. ESBMC-PLC supported the textual format but parsed graphical exports from CONTROLLINO, Beremiz, and OpenPLC Editor into an empty GOTO intermediate representation, causing vacuous verification success. This paper presents Graph-ESBMC-PLC, which closes this gap with a DFS-based graphical LD resolver. The resolver traverses the connection graph from leftPowerRail to each coil, extracts rung paths as Boolean contact conjunctions, and applies a three-tier I/O inference scheme. Ordering coils by rightPowerRail connectionPointIn sequence ensures SET coils process before RESET coils, matching IEC scan-cycle semantics. The graphical-to-IR conversion leaves the ESBMC backend unchanged. Validation on 3 graphical LD programs from CONTROLLINO/OpenPLC Editor shows all produce full GOTO IR with nondeterministic inputs and rung logic, versus the empty IR previously. All 3 verify SAFE at k=2 under 70ms. The 11 textual LD benchmarks are fully preserved, with no regression. Two Beremiz examples with no LD content or unsupported timer semantics are reported as discovered limitations. Artifact at Zenodo (DantasCordeiro2026graphical, doi:10.5281/zenodo.20699856).

04.
arXiv (CS.LG) 2026-06-11

Renewable Lasso without Batch-Number Constraints: A Gradient-Enhanced Approach

arXiv:2606.11738v1 Announce Type: cross Abstract: We study online estimation for high-dimensional generalized linear models with streaming data. First, for the non-distributed setting, we propose a gradient-enhanced surrogate loss that approximates the cumulative loss using only historical summaries, which modifies and improves upon the existing renewable estimation approach for the same model in the high-dimensional setting, and removes the batch-number constraint in previous studies. We then extend the method to distributed streaming data under the master-client architecture, where batches are partitioned across sites and only summaries (gradient vectors) are exchanged. Instead of directing applying the popular method of Jordan et al. (2019) to the surrogate quadratic loss, our adjusted approach does not require the clients to compute the full surrogate loss. We derive non-asymptotic error bounds under the high-dimensional scaling, without the stringent constraint on the number of batches in the previous studies. Simulation results under linear and logistic models, together with a real-data application, show improved accuracy over existing renewable estimators.

05.
arXiv (CS.LG) 2026-06-11

The ASE-LSE Disagreement Landscape: An End-to-End Characterisation of Extremes and Structural Drivers

arXiv:2605.22346v3 Announce Type: replace-cross Abstract: Two of the most widely used methods for analysing graph data, Adjacency Spectral Embedding and Laplacian Spectral Embedding, often produce different results when applied to the same graph. Yet the structural reasons behind this disagreement remain incompletely understood. This paper provides an end-to-end account of ASE-LSE latent subspace disagreement. We first prove that the two methods produce identical latent subspaces for every embedding dimension whenever the Laplacian is a scalar multiple of the adjacency matrix, and show that this scalar relationship holds if and only if the graph is either regular or bipartite biregular. This anchor result identifies a sufficient condition for perfect agreement that pins down the floor of the disagreement spectrum and supplies the baseline for the perturbation analysis. We then prove that no maximal-disagreement graph or family of graphs exists: the disagreement is always strictly below its theoretical ceiling, and we exhibit a witness family demonstrating that no finite maximum is attainable, so the disagreement landscape has no maximiser. With both endpoints established, we derive a Regularity Departure Bound whose two terms isolate degree heterogeneity and eigengap as the primary structural factors influencing disagreement in the middle regime. Empirical validation across thousands of simulated graphs confirms the mechanisms predicted by the bound: heterogeneity pushes disagreement up, eigengap suppresses it, and their joint ratio emerges as a unified predictor of ASE-LSE disagreement, suggesting when the two embeddings can be treated as interchangeable and when they cannot.

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

Geometry of Reason: Spectral Signatures of Valid Mathematical Reasoning

Verifying whether a language model is genuinely reasoning or pattern-matching remains an open problem: learned verifiers are expensive, and output-based heuristics are brittle. We show that valid mathematical reasoning induces a measurable, training-free spectral signature in transformer attention. By treating each attention matrix as a weighted token graph, we extract four diagnostics: Fiedler value, High-Frequency Energy Ratio (HFER), spectral entropy, and smoothness, that require no learned parameters. Experiments across seven models from four architectural families yield effect sizes up to Cohen's $d = 3.30$ ($p < 10^{-116}$), enabling $85$–$96\%$ single-threshold classification accuracy. Two findings sharpen the interpretation. First, Platonic validity: the spectral signal tracks logical coherence rather than compiler acceptance, proofs rejected for timeouts or missing imports are correctly classified as valid, a distinction confirmed by a manual audit ($\kappa = 0.82$, $n = 51$). Second, architectural determinism: Sliding Window Attention shifts the discriminative feature from HFER to smoothness ($d = 2.09$, $p < 10^{-48}$), showing that attention design governs which spectral channel encodes reasoning quality. Causal ablation confirms the signature traces induction-head circuits. The method generalises to informal chain-of-thought ($d = 0.78$, $p < 10^{-3}$), and in proof search, HFER reranking improves Best-of-16 Pass@1 by $+4.4$–$6.6$\%, matching $98\%$ of the AUC of fully supervised probes with zero labels. Spectral graph analysis is a principled, architecture-aware primitive for reasoning verification.

07.
arXiv (CS.AI) 2026-06-19

SARLO-80: Worldwide Slant SAR Language Optic Dataset 80cm

arXiv:2606.20523v1 Announce Type: cross Abstract: Multimodal foundation models have advanced rapidly thanks to large optical benchmarks, but comparable resources for synthetic aperture radar (SAR) remain limited. Existing SAR–optical datasets largely rely on low-resolution, intensity-only Ground Range Detected~(GRD) products and do not preserve complex-valued SAR measurements or native acquisition geometry, which restricts physically grounded multimodal learning. In particular, large-scale public datasets combining very-high-resolution (VHR) SAR SLC, aligned optical imagery, and natural-language descriptions are still lacking. We present a VHR SAR–optical–text dataset built from open-access Umbra spotlight acquisitions distributed as Sensor Independent Complex Data (SICD). From around 2,500 worldwide scenes (VV/HH, 20cm–2m native resolution), we standardize all SAR data to an 80cm slant-range grid via band-limited FFT resampling and tile the imagery into 1024 by 1024 patches. For each SAR patch, we retrieve a high-resolution optical tile and warp it into the SAR grid using local coordinate correspondences for local pixel-level alignment. We further generate three caption variants (SHORT/MID/LONG) per sample to support vision–language training and evaluation. Our dataset contains 119,566 triplets (complex and amplitude slant-range SAR patch, aligned optical patch, natural-language description) covering 257 locations across 72 countries and a broad range of land types and infrastructures. We release fixed train/validation/test splits and the full preprocessing and baseline code to enable reproducible benchmarks for multimodal alignment on cross-modal retrieval and conditional generation in native SAR geometry. The dataset is publicly available on the Hugging Face Hub at https://huggingface.co/datasets/ONERA/SARLO-80.

08.
arXiv (CS.CL) 2026-06-24

Self-Recognition Finetuning can Prevent and Reverse Emergent Misalignment

Emergent misalignment (EM) has been linked to the activation of misaligned persona vectors and evil character traits, suggesting that EM operates through disruption of the model's aligned character rather than direct learning of harmful content. Motivated by this connection, we study self-generated text recognition (SGTR) finetuning as a character-targeted intervention that is distinct from existing in-training defenses. We conduct two-stage finetuning experiments across three models (GPT-4.1, Qwen2.5-32B-Instruct, Seed-OSS-36B-Instruct) and multiple EM datasets to compare SGTR finetuning against benign finetuning baselines (correct domain-specific data, general knowledge, and word counting) to find it an effective defense in both reversal and prevention settings. We find that all interventions produce comparable EM reversal, but only when restoring capabilities that EM had degraded. For prevention, only SGTR finetuning consistently reduces misalignment without exacerbating any individual metric, suggesting that character fortification specifically drives prevention. We provide further evidence for EM's relation to the LLM's default character by showing that EM finetuning induces diversity into the LLM's identity self-reports, artificially corrupting self-recognition exacerbates misalignment caused by EM finetuning, and that removing the model's identity-bearing system prompt substantially reduces the effect of EM finetuning. Together, these findings reframe EM not as the adoption of a coherent misaligned persona but as the destabilization of aligned character.

10.
arXiv (CS.CL) 2026-06-16

DEEPRUBRIC: Evidence-Tree Rubric Supervision for Efficient Reinforcement Learning of Deep Research Agents

Deep research agents synthesize long-form reports by searching and reasoning over retrieved evidence. Reinforcement learning with rubric-based rewards improves these agents by optimizing them against checkable criteria that translate report quality into reward signals, but its efficiency depends on whether those criteria reliably capture the task scope and evidence needs. Most existing studies ask an LLM to generate rubrics for a given query, but when the model fails to infer the underlying information needs, the generated rubrics may be incomplete and reduce RL efficiency. To obtain more reliable query–rubric supervision, we introduce DeepRubric, a data construction framework that reverses this process: instead of inferring evaluation criteria for a given query, it first determines what an evidence-backed report should be evaluated on and then synthesizes aligned query–rubric pairs from those evaluation targets. Starting from a sampled seed topic, DeepRubric builds an evidence tree by recursively expanding evidence-backed sub-questions, whose leaves serve as atomic and verifiable evaluation targets. It then uses the evidence tree to synthesize the training query and rubrics, ensuring that the reward evaluates exactly the information requested by the query. Using DeepRubric, we construct 9K query–rubric supervision examples and train DeepRubric-8B with rubric-based GRPO, achieving comparable performance to prior open state-of-the-art deep research models across three benchmarks with roughly 13x fewer RL GPU-hours.

11.
arXiv (CS.CL) 2026-06-19

Segment-Level Mandarin Chinese Speech-Based Cognitive Impairment Detection via an Autoencoder with Contrastive Learning

\noindentBackground and Objective: Speech has emerged as a low-cost and non-invasive digital biomarker with considerable potential for cognitive impairment detection. However, limited labeled data and cross-dataset variability remain major challenges for robust speech-based screening systems. \par\noindentMethods: We developed a segment-level representation learning framework for speech-based cognitive impairment detection. Speech recordings were divided into short segments and converted into spectrogram representations. To improve robustness under limited-data conditions, offline and online augmentation strategies were combined with autoencoder-based representation learning and contrastive objectives to enhance discriminative latent representations. \par\noindentResults: Experiments conducted on four independent Mandarin Chinese speech datasets demonstrated stable and competitive performance in both binary and three-class classification tasks, with particularly notable improvements in the clinically challenging three-class setting. Ablation studies further supported the effectiveness of the proposed framework. \par\noindentConclusions: The findings suggest that segment-level speech representation learning may provide a scalable and practical approach for cognitive impairment screening in resource-constrained clinical settings.

12.
arXiv (quant-ph) 2026-06-11

The quantum harmonic oscillator and the real Hilbert space

arXiv:2606.12060v1 Announce Type: new Abstract: The harmonic oscillator is considered within generalized frameworks using complex and quaternionic numbers. The classical oscillator is considered in terms of a complex position function, and quantum oscillators are examined in terms of complex wave functions, and in terms of quaternionic wave functions as well. Both of the quantum solutions are obtained within the real Hilbert space formalism. The results reveal the complex and quaternionic descriptions as suitable frameworks for non-stationary processes, including damped oscillations, forced oscillations, and additionally self-interacting processes that cannot be appropriately described otherwise.

13.
arXiv (quant-ph) 2026-06-15

Emission of time-ordered photon pairs from a coherently-driven Kerr microcavity

arXiv:2601.06468v2 Announce Type: replace-cross Abstract: Weakly-interacting many-body systems possess remarkable quantum properties that are essential components of quantum technologies, and constitute a topic of fundamental interest. Here we show that in a solid-state nonlinear microcavity embedding discrete modes of exciton-dressed photons, we can isolate a single eigenmode of quantum fluctuations from the much brighter coherent fraction of the field. In this regime, we perform frequency- and time-resolved correlations measurements between photons on the red and blue side of the fluctuations spectrum. When the average number of fluctuation quanta is smaller than one, we observe the formation of large pairwise time-ordered correlations: red photon first and blue photon second. We show that this peculiar time-ordering correlation emerges spontaneously from the interplay between frequency-resolved detection, and the non-trivial internal quantum structure of the elementary fluctuations.

14.
arXiv (CS.CL) 2026-06-15

Can Post-Training Turn LLMs into Good Medical Coders? An Empirical Study of Generative ICD Coding

Automated International Classification of Diseases (ICD) coding is a core medical-coding task for billing, epidemiology, and clinical decision support. Generative large language models (LLMs) are often reported as weak medical coders, but this finding mainly comes from inference-time settings such as prompting, retrieval, reranking, or tool use, leaving the role of task-specific post-training underexplored. We present a controlled empirical study of post-training for generative ICD coding, comparing discriminative baselines with LLM coders across prompting, supervised fine-tuning, and reinforcement learning under a common protocol and metric set. To our knowledge, this is the first study to evaluate RL-based post-training for generative LLM coders in ICD coding. We further introduce PHI, a diagnostic curriculum that extends GRPO to refine missed-code cases. Our results show that prompting-only evaluation substantially underestimates the potential of LLMs for ICD coding. SFT provides the main capability jump, GRPO further improves code-set prediction beyond SFT, and PHI provides targeted gains on macro-level performance. These findings suggest that the main bottleneck is not the generative formulation alone, but how the model is adapted and optimized for full-taxonomy recall. We release our code, data splits, and checkpoints at https://github.com/AlexandreWANG915/LLM4ICD.

15.
arXiv (CS.CL) 2026-06-12

X-MADAM-RAG: Diagnosing and Handling Chinese-English Evidence Conflict in Retrieval-Augmented Generation

Retrieval-augmented generation (RAG) systems may receive evidence that is not merely noisy but mutually contradictory. This issue becomes particularly salient in multilingual settings, where retrieved Chinese and English evidence may support incompatible answer candidates. We study this problem through X-RAMDocs-ZHEN, a controlled Chinese-English benchmark derived from RAMDocs for diagnosing evidence conflict in RAG. The benchmark contains 300 examples across six balanced conditions, including monolingual support, bilingual agreement, reversed conflict directions, and conflict with optional noise. We further examine X-MADAM-RAG, an interpretable pipeline that decomposes evidence handling into per-document candidate extraction, visible-evidence repair, deterministic candidate grouping, and conflict-aware aggregation. On the original controlled benchmark with Qwen2.5-7B-Instruct, X-MADAM-RAG achieves 0.9667 strict accuracy and 0.9767 conflict-aware success, outperforming an evidence-normalized single-call baseline. However, a zero-call rule-only extractor reaches 1.0000 on the same benchmark, revealing strong template regularity. To probe this limitation, we construct a deterministic naturalized stress test that removes explicit answer templates while preserving candidate strings. On its 100-sample subset, rule-only extraction falls to 0.0000, but X-MADAM-RAG also drops to 0.3000 strict accuracy, below both naive and evidence-normalized baselines. A privileged oracle remains perfect, indicating that document-level extraction is the main bottleneck. These findings position X-RAMDocs-ZHEN and X-MADAM-RAG as diagnostic tools for controlled evidence conflict rather than as evidence of general hallucination detection or robustness to natural retrieval.

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

Towards Advanced Mathematical Reasoning for LLMs via First-Order Logic Theorem Proving

Large language models (LLMs) have shown promising first-order logic (FOL) reasoning capabilities with applications in various areas. However, their effectiveness in complex mathematical reasoning involving multi-step FOL deductions is still under-researched. While LLMs perform competitively on established mathematical reasoning benchmarks, they struggle with multi-step FOL tasks, as demonstrated by Deepseek-Prover-V2-7B's low accuracy (4.2%) on our proposed theorem proving dataset. This issue arises from the limited exploration of diverse proof strategies and the potential for early reasoning mistakes to undermine entire proofs. To address these issues, we propose DREAM, a self-adaptive solution that enhances the Diversity and REAsonability of LLMs' generation strategies. DREAM incorporates an Axiom-Driven Strategy Diversification mechanism to promote varied strategic outcomes and a Sub-Proposition Error Feedback to help LLMs reflect on and correct their proofs. Our contributions include pioneering advancements in LLMs' mathematical reasoning through FOL theorem proving, introducing a novel inference stage solution that improves performance by 0.6% to 6.4%, and providing a curated dataset of 447 mathematical theorems in Lean 4 format for evaluation.

17.
Nature (Science) 2026-06-10

Lignin to adipic acid in a high-yield chemical and biological redox process

Viable manufacturing pathways to produce bio-based chemicals from renewable feedstocks, such as lignin derived from plant biomass, are needed to decarbonize the chemicals manufacturing sector. Converting the recalcitrant lignin polymer to valuable bioproducts remains a longstanding challenge in biorefining, with the highest reported single-product yield from lignin currently around 20 wt% (refs. 1–4). Most existing lignin depolymerization strategies target aryl–ether bond cleavage, which can produce aromatic monomers in yields of only about 30 wt%, and&nbsp;still as complex mixtures with C–C-linked dimers and oligomers5,6. The recalcitrance of these C–C linkages between aromatic moieties fundamentally limits single-product yields from lignin, prompting the development of strategies to efficiently cleave these C–C bonds3,7–9. Here we show how reductive processing of lignin from poplar accesses a hydrocarbon mixture of alkyl-aromatic monomers and oligomers that is privileged for oxidative conversion to monomeric aromatic carboxylic acids, comprising mostly benzoic acid and phthalic acid isomers in up to 73 wt% monomer yields, using a Co/Mn/Br catalyst. The soil bacterium Pseudomonas putida KT2440 was engineered to convert this mixture of aromatic carboxylic acids to muconolactone, a precursor to bio-based nylons, enabling final adipic acid yields up to 26 wt% (gram adipic acid per gram lignin) with a maximum theoretical yield of 57&nbsp;wt%. This pairing of reductive and oxidative steps with lignin resembles processes in petrochemical refining and shows how lignin may be converted into a single, valuable bioproduct in high yields. A chemical and biological redox process that resembles processes in petrochemical refining is used to convert lignin from poplar into a single, valuable bioproduct, adipic acid, in high yields.

18.
arXiv (CS.AI) 2026-06-15

An Analysis of the Coordination Gap between Joint and Modular Learning for Job Shop Scheduling with Transportation Resources

arXiv:2604.24117v2 Announce Type: replace Abstract: Efficient job-shop scheduling with transportation resources is critical for high-performance manufacturing. With the rise of "decentralized factories", multi-agent reinforcement learning has emerged as a promising approach for the combined scheduling of production and transportation tasks. Prior work has largely focused on developing novel cooperative architectures while overlooking the question of when joint training is necessary. Joint training denotes the simultaneous training of job and automatic guided vehicle scheduling agents, whereas modular training involves independently training each agent followed by post-hoc integration. In this study, we systematically investigate the conditions under which joint training is essential for optimal performance in the job-shop scheduling problem with transportation resources. Through a rigorous sensitivity analysis of resource scarcity and temporal dominance, we quantify the coordination gap – the performance difference between these two training modalities. In our evaluation, joint training outperforms the majority of dispatching rule combinations and modular training approaches. However, the coordination gap advantage diminishes in bottleneck environments, particularly under severe transport and processing constraints. These findings indicate that modular training represents a viable alternative in environments where a single scheduling task dominates. Overall, our work provides practical guidance for selecting between training modalities based on environmental conditions, enabling decision-makers to optimize reinforcement learning-based scheduling performance.

19.
arXiv (CS.AI) 2026-06-16

AI Contagion in Social Networks

arXiv:2606.15206v1 Announce Type: cross Abstract: We study how artificial intelligence (AI) interacts with social communication networks to shape the stability of collective knowledge. Agents exchange information through a network while receiving AI-generated content, and AI systems retrain on the aggregate social information they influence. This interaction generates two feedback forces: an AI contagion channel, through which distortions diffuse across the network, and an AI social distortion multiplier, through which retraining amplifies past errors. Despite the high dimensionality of the environment, we show that the long-run behavior of the system admits a two-dimensional representation whose spectral radius determines whether AI-mediated information systems are dynamically stable or unstable. We characterize a sharp regulatory frontier identifying the minimum filtering required for stability and show how network topology shapes systemic informational risk.

20.
arXiv (CS.AI) 2026-06-11

From Architecture to Output: Structural Origins of Hallucination in Large Language Models and the Amplifying Role of Data

arXiv:2606.07537v1 Announce Type: cross Abstract: Large language models hallucinate–producing fluent, confident, factually wrong outputs–with a consistency that persists across generations and scales. Existing taxonomies classify hallucination by output type, distinguishing intrinsic from extrinsic failures and faithfulness from factuality divergence. These frameworks are descriptively rigorous but do not identify which internal mechanism produced a given instance. This paper analyses hallucination as a structural consequence of three architectural decisions that together form a compound failure system. Self-attention's co-occurrence learning substitutes statistical proximity for semantic meaning and produces entity confusion, fact misattribution, and semantic drift. The maximum likelihood estimation training objective optimises next-token probability without factual constraint, rewarding statistically plausible outputs regardless of their truth value. Autoregressive decoding's permanent left-to-right commitment under exposure bias ensures that a single wrong token cascades forward through the entire output sequence without revision. Dataset pathologies–long-tail deficiencies, training bias, and synthetic pollution–amplify these vulnerabilities but do not independently cause them. We make three contributions. First, we map each mechanism to a specific output category in the Alansari and Luqman taxonomy, locating intrinsic hallucination in self-attention, extrinsic hallucination in MLE, and logical inconsistency in autoregressive decoding. Second, we show that each commonly cited dataset pathology exploits one of these mechanisms rather than originating hallucination independently. Third, we identify the diagnostic limitation of output-type-only classification and contrast it with inference-layer mitigation approaches.

21.
arXiv (quant-ph) 2026-06-24

Erasure cost of a quantum process: A thermodynamic meaning of the dynamical min-entropy

arXiv:2506.05307v5 Announce Type: replace Abstract: The erasure of information is fundamentally an irreversible logical operation, carrying profound consequences for the energetics of computation and information processing. We investigate the thermodynamic costs associated with erasing (and preparing) quantum processes. Specifically, we analyze an arbitrary bipartite unitary gate acting on logical and ancillary input-output systems, where the ancillary input is always initialized in the ground state. We focus on the adversarial erasure cost of the reduced dynamics - that is, the minimal thermodynamic work cost to erase the logical output of the gate for any logical input, assuming full access to the ancilla but no access to any purifying reference of the logical input state. We determine that this adversarial erasure cost is directly proportional to the negative min-entropy of the reduced dynamics, thereby giving the dynamical min-entropy a clear operational meaning. The dynamical min-entropy can take positive and negative values, depending on the underlying quantum dynamics. The negative value of the erasure cost implies that the extraction of thermodynamic work is possible instead of its consumption during the process. A key foundation of this result is the quantum process decoupling theorem, which quantitatively relates the decoupling ability of a process with its min-entropy. This insight bridges thermodynamics, information theory, and the fundamental limits of quantum computation.

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

Dissociating Decodability and Causal Use in Bracket-Sequence Transformers

When trained on tasks requiring an understanding of hierarchical structure, transformers have been found to represent this hierarchy in distinct ways: in the geometry of the residual stream, and in stack-like attention patterns maintaining a last-in, first-out ordering. However, it remains unclear whether these representations are causally used or merely decodable. We examine this gap in transformers trained on the Dyck language (a formal language of balanced bracket sequences), where the hierarchical ground truth is explicit. By probing and intervening on the residual stream and attention patterns, we find that depth, distance, and top-of-stack signals are all decodable, yet their causal roles diverge. Specifically, masking attention to the true top-of-stack position causes a sharp drop in long-distance accuracy, while ablating low-dimensional residual stream subspaces has comparatively little effect. These results, which extend to a templated natural language setting, suggest that even in a controlled setting where the relevant hierarchical variables are known, decodability alone does not imply causal use.

23.
arXiv (CS.LG) 2026-06-16

Multi-User Dueling Bandits: A Fair Approach using Nash Social Welfare

arXiv:2605.01961v2 Announce Type: replace Abstract: Learning from human preference data is becoming a useful tool, from fine-tuning large language models to training reinforcement learning agents. However, in most scenarios, the model is trained on the average preference of all human evaluators, which, under large variations of preferences, can be unfair to minority groups. In this work, we consider fairness in dueling bandits, a standard framework for online learning from preference data. We assume that each user has a (potentially distinct) Condorcet winner, which is an arm preferred to every other arm. Using these user-specific Condorcet winners as reference points, we evaluate and score arms according to their performance relative to the corresponding winner. To promote fairness across heterogeneous users, we adopt the well-established Nash Social Welfare objective, which maximizes the product of user utilities, thereby inherently penalizing inequality and preventing the marginalization of any single user. Within this framework, we construct a hard instance to establish a regret lower bound of $\Omega(T^{2/3}\min(K,D)^\frac{1}{3})$ for a time horizon $T$, $K$ arms, and $D$ users, which, to the best of our knowledge, is the first result quantifying the cost of fairness in dueling bandits with heterogeneous preferences. We then present the Fair-Explore-Then-Commit and Fair-$\epsilon$-Greedy algorithms with a Condorcet winner identification phase. We further derive their regret upper bounds that match the lower-bound dependence on $T$ up to logarithmic factors.

24.
arXiv (CS.AI) 2026-06-16

The Missing Knowledge Layer in Cognitive Architectures for AI Agents

arXiv:2604.11364v2 Announce Type: replace Abstract: The two most influential cognitive architecture frameworks for AI agents, CoALA [21] and JEPA [12], both lack an explicit Knowledge layer with its own persistence semantics. This gap produces a category error: systems apply cognitive decay to factual claims, or treat facts and experiences with identical update mechanics. We survey persistence semantics across existing memory systems and identify eight convergence points, from Karpathy's LLM Knowledge Base [10] to the BEAM benchmark's near-zero contradiction-resolution scores [22], all pointing to related architectural gaps. We propose a four-layer decom position (Knowledge, Memory, Wisdom, Intelligence) where each layer has fundamentally different persistence semantics: indefinite supersession, Ebbinghaus decay, evidence-gated revision, and ephemeral inference respectively. Companion implementations in Python and Rust demonstrate the architectural separation is feasible. We borrow terminology from cognitive science as a useful analogy (the Knowledge/Memory distinction echoes Tulving's trichotomy), but our layers are engineering constructs justified by persistence-semantics requirements, not by neural architecture. We argue that these distinctions demand distinct persistence semantics in engineering implementations, and that no current framework or system provides this.

25.
arXiv (CS.AI) 2026-06-18

Learning from Own Solutions: Self-Conditioned Credit Assignment for Reinforcement Learning with Verifiable Rewards

arXiv:2606.18810v1 Announce Type: cross Abstract: Reinforcement learning with verifiable rewards (RLVR) has driven substantial progress in training LLMs for reasoning tasks, but representative methods such as GRPO assign uniform credit across all tokens, wasting gradient on routine tokens while under-crediting pivotal reasoning steps. Existing token-level credit assignment methods require resources beyond the model's own rollouts. GRPO variants rely on process reward models or ground-truth answers. Knowledge distillation assigns credit through per-token divergence but requires external teachers (On-Policy Distillation) or privileged information (On-Policy Self Distillation). However, these dependencies limit applicability in the pure RLVR setting. We observe that conditioning the model on its own verified trajectories induces a measurable per-token KL divergence between the original and conditioned distributions, and prove that distilling from a self-teacher constructed by verified trajectories leads to infeasible weighted-average solutions when multiple verified trajectories exist. We propose SC-GRPO (Self-Conditioned GRPO), which uses KL divergence mentioned before as a multiplicative weight on GRPO gradients. Across five benchmarks spanning math, code, and agentic tasks, SC-GRPO consistently outperforms 8.1% over GRPO and 5.9% over DAPO with stronger OOD performance. Moreover, SC-GRPO achieves higher performance than OPD.