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

Benchmarking Agentic Review Systems

A new class of agentic review systems are emerging as a remedy to the pressure placed on peer review systems by AI-assisted research, but it is unclear how they should be evaluated. We evaluate two open-source systems (OpenAIReview and coarse), one proprietary system (Reviewer3), and a zero-shot baseline, across six LLMs spanning frontier and efficient models. First, we study whether AI reviews on ICLR/NeurIPS papers track with papers' quality as approximated by external signals such as citations and acceptance decisions. Every system performs above chance in pairwise accuracy, and the best is OpenAIReview + GPT-5.5 at 83.0%. Second, to test whether systems can catch errors with known ground truth, we construct a perturbation benchmark that injects four categories of errors into papers across eight arXiv subject classes and measure detection recall. The strongest configuration (OpenAIReview + GPT-5.5) catches 71.6% of injected errors, leaving substantial room for improvement. The union of detections across six models reaches 83.3% recall, suggesting different models detect different errors and better harness design can potentially increase performance. Beyond these benchmarks, we study a public deployment of OpenAIReview with real users. Votes on its comments skew positive at 1.44 to 1, and the most common complaints are about false positives and minor nitpicks. Together, by evaluating full review systems backed by state-of-the-art models on real research papers, we show that while AI reviews still have room for improvement, they can already track human quality judgments well, catch important errors, and earn positive feedback from real users.

02.
arXiv (CS.CV) 2026-06-17

ERQA-Plus: A Diagnostic Benchmark for Reasoning in Embodied AI

Generalist embodied agents require more than object recognition: they must reason about spatial relations, actions, procedures, human intentions, environmental constraints, and commonsense consequences from situated visual observations. Yet existing visual and embodied question answering benchmarks often provide limited control over the reasoning dependencies being tested, making it difficult to distinguish grounded embodied reasoning from shortcut-driven visual or linguistic pattern matching. We present ERQA-Plus, a diagnostic benchmark for reasoning in embodied AI. ERQA-Plus contains 1,766 question-answer instances grounded in 711 robot-centric images and organized according to a structured taxonomy spanning perceptual, action-centric, social-interaction, navigation-environmental, and contextual commonsense reasoning. The dataset is constructed using a multi-stage generation and validation pipeline that combines taxonomy-guided question generation, automatic quality judging, iterative revision, and human assessment to improve visual grounding, answer validity, and reasoning quality. We benchmark representative general-purpose vision-language models and embodied models, including LLaVA-NeXT-8B, Prismatic-7B, MiniCPM-V-4.5-8B, Qwen3-VL, RoboRefer-8B, and RoboBrain2.5-8B. Although the strongest model, Qwen3-VL-32B, achieves 83.4% overall accuracy and 61.4 SBERT score, category-level results reveal persistent weaknesses in spatial reasoning, procedural reasoning, event prediction, and intention inference. ERQA-Plus therefore provides a fine-grained evaluation framework for measuring not only whether embodied agents answer correctly, but also which forms of embodied reasoning they can and cannot perform reliably. The dataset is available https://huggingface.co/datasets/huggingdas/erqa-plus and the project page at https://github.com/LUNAProject22/erqa-plus.

03.
arXiv (CS.LG) 2026-06-12

Let's Ask Gauss: Improved One-Run Privacy Auditing

arXiv:2606.12733v1 Announce Type: new Abstract: Privacy auditing provides an important safeguard by estimating the actual information leaked by a model, thus ensuring that theoretical privacy guarantees hold in practice. We study empirical privacy auditing for differentially private (DP) machine learning, focusing on efficient one-run methods for mechanisms such as DP-SGD. Prior one-run approaches threshold training examples or "canaries" into binary membership guesses, which discards useful information. We show that, in the white-box DP-SGD setting, canary-aligned signals naturally form a sequence of random variables whose normalized sum is asymptotically Gaussian. Leveraging this distributional perspective, we develop a DP-auditing framework that leads to tighter privacy lower bounds from a single training run.

04.
Nature Biotechnology 2026-06-05

Structural motif search across the protein universe with Folddisco

作者:

Detecting similar protein structural motifs in large structure collections is computationally expensive. We developed Folddisco, a fast structural motif search tool that uses an index of position-independent geometric features, including side-chain orientation, combined with a rarity-based scoring system. Folddisco is 20-fold faster in querying and fourfold more storage-efficient than existing methods while improving accuracy. Folddisco is freely available online ( https://folddisco.foldseek.com ), along with a webserver ( https://search.foldseek.com/folddisco ). Folddisco enables protein structural motif search in million scale databases.

05.
arXiv (CS.CL) 2026-06-12

SENTINEL: Failure-Driven Reinforcement Learning for Training Tool-Using Language Model Agents

Language model agents are increasingly effective in solving realistic tasks through multi-turn tool use. However, training reliable tool-using agents remains challenging in practice. While reinforcement learning provides an on-policy paradigm for improving agents from their own environment interactions, its effectiveness depends heavily on the training task distribution. When tasks are fixed before training, the task distribution can become increasingly mismatched with the policy's evolving capabilities, causing many rollouts to be spent on uninformative tasks. We propose SENTINEL, a failure-driven reinforcement learning framework that turns the Solver's rollout failures into targeted training tasks. SENTINEL follows a Controller–Proposer–Solver loop: the Controller analyzes failed trajectories and summarizes recurring error patterns, the Proposer generates executable tasks that stress these weaknesses, and the Solver is trained on the targeted tasks. On Tau2-Bench Retail with Qwen3-4B-Thinking-2507, SENTINEL improves Pass\^{}1 from 66.4 to 74.9 and outperforms RL on general synthetic tasks across Pass\^{}k metrics. These results demonstrate that model failures provide an effective and scalable source of targeted training signal for improving tool-using language model agents.

06.
arXiv (CS.CV) 2026-06-19

CARE: Competence-Aware Reward Shaping for Adaptive Reasoning Length in Video-MLLMs

In multimodal video reasoning, reinforcement learning-based methods typically rely on simplistic and inflexible reasoning-length control strategies that fail to adapt to the model's evolving competence. This mismatch may suppress necessary exploration at early stages, while encouraging redundant reasoning and inefficient decoding once the model becomes more competent. In this paper, we propose CARE, a competence-aware reward shaping framework for adaptive reasoning length optimization in multimodal reasoning. Specifically, CARE maintains a smoothed competence estimate via an exponential moving average of pass rates, and uses it to route training into progressive stages that shift the reward preference from exploration-oriented long-form reasoning to efficiency-oriented concise reasoning. To avoid conflating verbosity with intrinsic task complexity, CARE further normalizes reasoning effort with batch-level statistics, and introduces a posterior amplifier to strengthen reward signals for unexpectedly strong performance on historically difficult samples. The proposed mechanism is seamlessly integrated into the GRPO training pipeline and incurs no additional inference-time overhead. Extensive experiments on multiple video reasoning and general video understanding benchmarks demonstrate that CARE consistently improves reasoning accuracy, stabilizes reinforcement learning, and significantly enhances token efficiency. Moreover, CARE exhibits a characteristic inverted-U trajectory of reasoning length during training, and yields shorter yet more informative reasoning traces at convergence, indicating effective adaptive allocation of reasoning budget. We provide the source code for our proposed CARE framework and experiments at https://github.com/1Pansy/Video-CARE.

07.
arXiv (CS.AI) 2026-06-17

Like a Hammer, It Can Build, It Can Break: Large Language Model Uses, Perceptions, and Adoption in Cybersecurity Operations on Reddit

arXiv:2604.09998v2 Announce Type: replace-cross Abstract: Large language models (LLMs) have recently emerged as promising tools for augmenting Security Operations Center (SOC) workflows, with vendors increasingly marketing autonomous AI solutions for SOCs. However, there remains a limited empirical understanding of how such tools are used, perceived, and adopted by real-world security practitioners. To address this gap, we conduct a mixed-methods analysis of discussions in cybersecurity-focused forums to learn how a diverse group of practitioners use and perceive modern LLM tools for security operations. More specifically, we analyzed 892 posts between December 2022 and September 2025 from three cybersecurity-focused forums on Reddit, and, using a combination of qualitative coding and statistical analysis, examined how security practitioners discuss LLM tools across three dimensions: (1) their stated tools and use cases, (2) the perceived pros and cons of each tool across a set of critical factors, and (3) their adoption of such tools and the expected impacts on the cybersecurity industry and individual analysts. Overall, our findings reveal nuanced patterns in LLM tools adoption, highlighting independent use of LLMs for low-risk, productivity-oriented tasks, alongside active interest around enterprise-grade, security-focused LLM platforms. Although practitioners report meaningful gains in efficiency and effectiveness in LLM-assisted workflows, persistent issues with reliability, verification overheads, and security risks sharply constrain the autonomy granted to LLM tools. Based on these results, we also provide recommendations for developing and adopting LLM tools to ensure the security of organizations and the safety of cybersecurity practitioners.

08.
arXiv (CS.CV) 2026-06-15

One Layer's Trash is Another Layer's Treasure: Adaptive Layer-wise Visual Token Selection in LVLMs

Large Vision-Language Models (LVLMs) have achieved remarkable success across diverse multimodal tasks, yet their practical deployment remains constrained by the computational burden arising from lengthy visual tokens. While visual token pruning has emerged as a promising solution, existing methods suffer from a fundamental limitation: once tokens are pruned at a specific layer, they become inaccessible to all subsequent layers, leading to premature information loss that can compromise model performance. Through empirical studies, we observe that different layers exhibit distinct visual region focus, indicating a varying optimal token subset across layers. Motivated by this insight, we propose Adaptive Layer-wise Visual Token Selection (ALVTS), a novel framework that breaks away from the conventional static token pruning paradigm. ALVTS incorporates a lightweight token selector to identify and route important tokens for further processing, while allowing less important tokens to skip the layer, thus minimizing computational redundancy. These two streams of tokens are seamlessly reintegrated before being fed into subsequent layers, facilitating adaptive compression across the entire model. Grounded in our importance consistency constrained low-rank approximation, the proposed token selection module closely emulates the full attention mechanism, effectively capturing its essential patterns without requiring model retraining. Extensive experiments on LLaVA-1.5, LLaVA-NeXT, and Qwen2.5-VL validate the effectiveness of our method. With an 89% token compression ratio, ALVTS retains 96.7% of the original model's accuracy, achieving a superior efficiency-accuracy trade-off for LVLM inference.

09.
arXiv (CS.AI) 2026-06-16

Deep Q-Learning on Hölder Spaces

作者:

arXiv:2606.16846v1 Announce Type: cross Abstract: We study the operator-theoretic core of Q-learning in continuous-time stochastic control with continuous states and actions. In value-based reinforcement learning, each Q-learning or DQN update is built from a Bellman optimality target; our analysis isolates this target in a diffusion setting and studies its regularity and approximation complexity. Under uniform ellipticity and Hölder-regular coefficients, we show that a Bellman update maps bounded inputs into an anisotropic regularity class, smoothing the state variable while leaving only Lipschitz dependence on the action variable. This yields a compact family of Bellman iterates and motivates a tensor-product DeepONet architecture adapted to the mixed regularity of the problem. We then derive explicit approximation and resource bounds, together with a stiffness–complexity trade-off as the time step $\delta \to 0$. The resulting theory makes a direct contribution to Q-learning theory at the level of Bellman target regularity and approximation in continuous stochastic control. At the same time, we do not claim a full convergence theorem for practical sampled Q-learning with exploration, replay, and stochastic gradient updates.

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

Dynestyx: A Probabilistic Programming Library for Dynamical Systems

arXiv:2606.16985v1 Announce Type: cross Abstract: State-space models (SSMs) are the standard formalism for Bayesian treatment of dynamical systems, with natural applications in statistics, signal processing, and machine learning. Despite their importance in both theory and application, dynamical systems have proven difficult to incorporate in modern probabilistic programming languages (PPLs), making state-of-the-art methods less accessible to practitioners and introducing friction in following the "Bayesian workflow." We introduce dynestyx, a probabilistic programming library with first-class support for SSMs, including state-of-the-art methods in the estimation of both states and parameters. Through a single, unified interface, users may specify arbitrary priors for discrete-time or continuous-time dynamical systems, perform inference over mixed-effect data, and make state and parameter estimates with principled uncertainty quantification.

11.
arXiv (CS.AI) 2026-06-15

The Accountability Paradox: How Platform API Restrictions Undermine AI Transparency Mandates

arXiv:2505.11577v5 Announce Type: replace-cross Abstract: Recent application programming interface (API) restrictions on major social media platforms challenge compliance with the EU Digital Services Act [20], which mandates data access for algorithmic transparency. We develop a structured audit framework to assess the growing misalignment between regulatory requirements and platform implementations. Our comparative analysis of X/Twitter, Reddit, TikTok, and Meta identifies critical ``audit blind-spots'' where platform content moderation and algorithmic amplification remain inaccessible to independent verification. Our findings reveal an ``accountability paradox'': as platforms increasingly rely on AI systems, they simultaneously restrict the capacity for independent oversight. We propose targeted policy interventions aligned with the AI Risk Management Framework of the National Institute of Standards and Technology [80], emphasizing federated access models and enhanced regulatory enforcement.

12.
arXiv (CS.AI) 2026-06-11

MPC-Patch-Bench: Security-Aware LLM Code Patch for Multi-Party Computation

arXiv:2606.11416v1 Announce Type: cross Abstract: Repository-level benchmarks for evaluating Large Language Model (LLM) code repair on Secure Multi-Party Computation (MPC) software do not yet exist, and directly transplanting general-purpose benchmarks such as SWE-bench fails on three structural fronts: (i) MPC repositories are dominated by generic Python infrastructure rather than cryptographic logic; (ii) high-value MPC fixes lack the standardized tests rigid extraction pipelines require; and (iii) standard fail-to-pass evaluation is insufficient for code that must also be cryptographically safe. MPC is increasingly deployed for privacy-preserving machine learning, biomedical collaboration, and secure analytics. Existing MPC-specific code-synthesis efforts cover only operator-level or single-framework tasks; evaluating LLM agents on real repository-level MPC repair instead demands MPC-aware data curation and a verifier matched to the security and numerical-fidelity guarantees MPC programs must obey neither of which existing benchmarks provide. We introduce MPC-Patch-Bench, a repository-level benchmark organised around two frameworks. (1)The Data Curation Framework combines a domain-specific curation agent that filters raw pull requests through three cryptographic layers with a human-AI completion engine that synthesizes missing problem statements and Fail-to-Pass/Pass-to-Pass tests, yielding 205 fully verified instances. (2)The MPC Verifier provides dedicated security and numerical-fidelity checks via dynamic differential testing against plaintext oracles and MPC-specific static analysis rules that flag unsafe reveals, insecure arithmetic, and illegal public/private casts. The strongest evaluated LLM functionally resolves only 22.9% of MPC-Patch-Bench tasks; the MPC Verifier further reduces verified resolution to 17.1%, with up to 40% of functionally-passing patches rejected for cryptographic or numerical-fidelity violations.

13.
arXiv (CS.CV) 2026-06-19

Rethinking Robust Adversarial Concept Erasure in Diffusion Models

Concept erasure aims to selectively unlearning undesirable content in diffusion models (DMs) to reduce the risk of sensitive content generation. As a novel paradigm in concept erasure, most existing methods employ adversarial training to identify and suppress target concepts, thus reducing the likelihood of sensitive outputs. However, these methods often neglect the specificity of adversarial training in DMs, resulting in only partial mitigation. In this work, we investigate and quantify this specificity from the perspective of concept space, i.e., can adversarial samples truly fit the target concept space? We observe that existing methods neglect the role of conceptual semantics when generating adversarial samples, resulting in ineffective fitting of concept spaces. This oversight leads to the following issues: 1) when there are few adversarial samples, they fail to comprehensively cover the object concept; 2) conversely, they will disrupt other target concept spaces. Motivated by the analysis of these findings, we introduce S-GRACE (Semantics-Guided Robust Adversarial Concept Erasure), which grace leveraging semantic guidance within the concept space to generate adversarial samples and perform erasure training. Experiments conducted with seven state-of-the-art methods and three adversarial prompt generation strategies across various DM unlearning scenarios demonstrate that S-GRACE significantly improves erasure performance 26%, better preserves non-target concepts, and reduces training time by 90%. Our code is available at https://github.com/Qhong-522/S-GRACE.

14.
arXiv (quant-ph) 2026-06-12

Influence-solvability: a systematic theory of $(1+1)D$ solvability and its application to brickwork circuits

arXiv:2606.12538v1 Announce Type: cross Abstract: `Solvable' circuits, such as dual unitaries and its generalisations, have arisen as paradigmatic examples of tractable chaotic non-equilibrium dynamics, both in classical and quantum systems. However, while increasingly more complicated sufficient conditions have been proposed, a systematic theory classifying and understanding general features of solvable circuits is missing. We develop such a theory by introducing influence-solvable circuits, a class of $(1+1)D$ circuits whose influence matrix, which represents the `bath' generated by its own evolution, is given by a uniform MPS with finite bond-dimension $\chi$. This property allows for efficient computation of subsystem dynamics and essentially contains all known examples of solvable circuits. We derive a set of necessary and sufficient local conditions by using a version of the fundamental theorem of MPS for open boundary conditions. Next we apply our theory to brickwork circuits with $\chi=1$ influence-solvability and perform a systematic classification of classical brickwork circuits with local dimension up to $d=3$ and quantum brickwork circuits with $d=2$. Our search reveals new solvable circuits that are not captured by known solvability conditions.

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

As Easy as Rocket Science: Assessing the Ability of Large Language Models to Interpret Negation in Figurative Language

Figurative language and negation are two areas that challenge current language models, however, both are widely used throughout written and spoken language. Large language models (LLMs) are also widely used in everyday contexts where they cannot necessarily be tuned for a specific dataset. It is therefore essential to understand the ability of LLMs to correctly interpret text that includes both negation and figurative language. To investigate this, we develop a set of new annotations to an existing dataset of figurative language, and test a range of language models on the dataset. We find that the combination of negation and figurativeness can present a particular challenge, and that performance overall and across different negation types is particularly dependent on the prompt style used.

16.
arXiv (CS.CV) 2026-06-17

UoU: A Universal Fingerprint Foundation Model Based on Large-Scale Unsupervised Learning

Fingerprint recognition is still dominated by task-specific pipelines, where enhancement, structural parsing, alignment, and matching are optimized in isolation. Although effective in narrow settings, this design limits representation reuse across sensors, qualities, and downstream applications. We therefore present UoU, short for ``a Universal fingerprint foundation model based on large-scale Unsupervised learning,'' which reframes fingerprint feature extraction as a domain-specific foundation-model problem. UoU is organized around a multi-level representation hierarchy spanning image restoration, structural fields, semantic tokens, point-level biometric entities, and compact global descriptors. Its training recipe combines a supervised cold start on precise annotations, large-scale weakly supervised refinement, and large-scale unsupervised consolidation, with the latter two stages iterated during large-scale training so that weak supervision broadens semantic coverage while unsupervised learning stabilizes correspondences, invariances, and representation geometry. Rather than treating fingerprint imagery as generic texture, UoU exploits domain-specific symmetries and intermediate structure, including orientation flow, periodic ridge patterns, sparse biometric entities, and spatial equivariance. The framework is intentionally architecture-agnostic: while the present study includes an initial transformer-based structured-prediction instantiation, the broader design supports multi-task learning, scalable model configurations, and downstream specialization for matching, alignment, enhancement, registration, and related fingerprint applications. This paper presents the technical motivation, system design, and validation protocol of UoU, and part of the baseline implementation is publicly available at https://github.com/XiongjunGuan/UoU.

17.
arXiv (CS.CL) 2026-06-16

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

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

18.
arXiv (CS.CV) 2026-06-17

TaFD: Threat-Aware Frequency Decoupling for Adversarial Robustness against Heterogeneous Attacks

Multi-threat robustness remains a fundamental challenge in deep learning. Although joint adversarial training (JAT) is widely adopted, it suffers from negative transfer under heterogeneous threats, particularly between $\ell_p$-bounded and semantic attacks. Through first-order gradient analysis, we formalize this as gradient incompatibility and theoretically establish the necessity of decoupled optimization. We further reveal that these conflicting threats exhibit separable spectral characteristics in the frequency domain. Motivated by this observation, we propose Threat-aware Frequency Decoupling (TaFD), a two-stage defense framework that reformulates JAT as a frequency-domain divide-and-conquer paradigm. TaFD first discovers latent threat domains via unsupervised clustering of attack spectral prototypes and trains a lightweight classifier for inference-time threat domain identification. Conditioned on the prediction, TaFD employs a Frequency-Conditional Convolution that learns threat-domain-specific spectral masks and routes each sample to the corresponding expert, enforcing structural parameter separation and alleviating optimization conflicts. We validate TaFD on three representative image-classification benchmarks (CIFAR-10, CIFAR-100, and Tiny-ImageNet) and on two representative architectures (the convolutional ResNet and the hybrid-transformer MobileViT). Extensive results demonstrate that TaFD achieves more balanced robustness against heterogeneous attacks than existing JAT and frequency-domain baselines, improving average robust accuracy by approximately 11\% over the strongest baseline while maintaining leading clean accuracy.

19.
arXiv (quant-ph) 2026-06-12

Metabolic quantum limit to the information capacity of magnetoencephalography

arXiv:2511.06401v3 Announce Type: replace-cross Abstract: Magnetoencephalography measures the magnetic fields generated by neural currents using quantum sensors such as superconducting quantum interference devices and atomic magnetometers. Here we combine the energy resolution limit of magnetic sensing with the metabolic power available to neural currents to derive a technology-independent bound on the information capacity of MEG. The bound factorizes into geometry, metabolism, and Planck's constant, and gives an estimated maximum information rate of 2.2~Mbit/s for representative human-brain parameters. Further, we show that the externally measurable magnetic field has a finite angular bandwidth, with high multipole components being geometrically attenuated and falling below the quantum-limited noise floor. This yields an information-limited spatial scale of order $1~cm$ and renders the accessible measurement space effectively finite-dimensional. The energy resolution limit therefore defines an information-theoretic Nyquist scale for magnetoencephalography, beyond which denser spatial sampling provides redundant measurements rather than additional recoverable information. Since the energy resolution limit also makes the noise variance grow linearly with measurement bandwidth, temporal and spatial bandwidths compete, producing a fundamental spatio-temporal trade-off. These results show how quantum-limited measurements constrain the observable complexity and information content of noninvasive brain imaging, providing a quantitative link between fundamental physics and neuroscience.

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

The Impossibility of Eliciting Latent Knowledge

arXiv:2606.12268v1 Announce Type: new Abstract: Advanced AI systems have extensive knowledge of their environments; in fact, their knowledge may (far) exceed that of their developers or users. Consequently, a desirable property for an AI system is that it is honest – that it accurately reports its beliefs about the world. Designing an AI system to be honest may be difficult, especially if we want to ask it questions about latent variables in the environment – variables which are hidden from the human interacting with it. This gives rise to the problem of eliciting latent knowledge (ELK): the problem of training an AI agent to honestly report its beliefs. In this paper, we make ELK formally precise using Causal Influence Diagrams (CIDs). CIDs can be used to describe the relationship between an agent's training environment and its subjective representation of the world. We use CIDs to formalise the distinction between observable and latent variables, to specify what exactly it means for an agent to be honest, and to formally define goal misgeneralisation. We show that, under certain circumstances, developers can incentivise an agent to honestly answer questions by providing correct feedback during training. However, a natural, but undesirable, way for an agent to generalise is to provide answers which humans would evaluate as true, rather than honest answers. We prove an impossibility theorem stating: There is no feedback-based training strategy that depends only on agent behaviour and with certainty produces an honest agent, even if feedback is perfect during training.

21.
arXiv (CS.AI) 2026-06-18

Attribution-Guided and Coverage-Maximized Pruning for Structural MoE Compression

arXiv:2606.18304v1 Announce Type: cross Abstract: Mixture-of-Experts (MoE) models scale compute efficiently, yet remain expensive to deploy due to their substantial memory footprint and inference overhead. Prior compression methods mainly operate at the expert level, either removing entire experts or ranking experts by coarse-grained importance scores. However, such expert-wise decisions are often too coarse to capture fine-grained redundancy, leading to misallocated pruning budgets and limited compression. To address this problem, we observe that information within MoE experts is highly concentrated in a small subset of channels, leaving substantial redundancy even in experts deemed important. Based on this observation, we propose a structural pruning framework tailored for MoE models. Our method reformulates prune-ratio allocation as a channel-score coverage maximization problem and solves it efficiently using an attribution-based approximation. Experiments on DeepSeek and Qwen MoE models show that our method preserves model accuracy under 50% or 25% structured pruning when combined with 4-bit quantization. On Qwen3-30B-A3B, our approach reduces memory footprint by 5.27$\times$ and consistently outperforms state-of-the-art baselines across diverse benchmarks.

22.
bioRxiv (Bioinfo) 2026-06-18

Calculation of sequence space coverage in a mutagenesis library

Directed evolution requires screening of large mutagenesis libraries, but accurate calculation of library sizes needed to discover functional variants remains challenging. Existing models provide baseline estimates, yet current computational approaches for finding the best variants scale poorly with library complexity. Here, we introduce a scalable algorithmic framework to compute exact discovery probabilities in saturation mutagenesis libraries with no requirement for explicit sequence enumeration. By aggregating variants into a composition log–sum distribution and applying log-space convolution across randomisation blocks, it is possible to extend this to massive sequence spaces and mixed codon schemes. By inverting these calculations, absolute mathematical ceilings for experimental design are established. Ultimately, this framework provides a rapid, quantitative tool to balance the statistical coverage-diversity trade-off within the limitations of laboratory screening. Finally, this is implemented as an open-source web application (SSCC) that allows researchers to construct heterogeneous library designs and compute required sampling depths, coverage probabilities, and absolute randomisation limits.

23.
arXiv (CS.CL) 2026-06-17

From Trainee to Trainer: LLM-Designed Training Environment for RL with Multi-Agent Reasoning

Reinforcement learning pipelines for Large Language Model (LLM) training often rely on manually redesigned environments between stages, requiring practitioners to heuristically infer which configuration will best improve the current policy. To automate this process, we propose the LLM-as-Environment-Engineer framework in which the current policy model analyzes failure trajectories together with contextual information and proposes modifications to the next-stage training environment configuration. We also introduce MAPF-FrozenLake, a controllable testbed whose generator exposes multi-dimensional environment configurations, making it suitable for studying and benchmarking environment redesign. On this testbed, we condition the environment engineer on structured summaries of policy behavior, failure cases, and environment statistics, from which it produces the configuration for the next training stage. With Qwen3-4B as the backbone, our framework achieves the strongest aggregate performance on our benchmarks, outperforming larger proprietary LLMs (e.g., GPT, Gemini) and fixed-environment training baselines. We further analyze which forms of context are most effective, finding that successful environment updates rely on failure evidence and preserve configurations that already work. Interestingly, the current RL checkpoint serves as a better environment engineer than the original base model, suggesting that policy learning improves the model's ability to diagnose its remaining weaknesses.

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

SGFormer++: Semantic Graph Transformer for Incremental 3D Scene Graph Generation

In this paper, we propose SGFormer++, a novel Semantic Graph Transformer for 3D scene graph generation (SGG), which aims to parse point cloud scenes into semantic structural graphs, where nodes denote detected object instances and edges encode their pairwise relationships, with the core challenge lying in modeling complex global scene structure. While existing graph convolutional network (GCN)-based methods suffer from over-smoothing and limited receptive fields, SGFormer++ leverages Transformer layers as its backbone to enable global message passing. Specifically, we introduce two key components tailored for 3D SGG: (1) a Graph Embedding Layer++ that efficiently integrates edge-aware global context with linear computational complexity, and (2) a Semantic Injection Layer++ that enriches visual features with linguistic priors from large language models (LLMs) and vision-language models (VLMs), boosting semantic representation without introducing extra trainable parameters. To further address the practical challenge of incremental SGG (I-SGG), where new relationship categories arrive sequentially, we equip SGFormer++ with a novel Spatial-guided Feature Adapter, which calibrates predicate features using subject-object spatial geometry to counter scale variation, and a Cascaded Binary Prediction Head that mitigates catastrophic forgetting via task-incremental classifier expansion and logit distillation. Extensive experiments on the 3DSSG benchmark demonstrate that SGFormer++ achieves state-of-the-art performance in both standard and incremental settings: it yields a significant 4.49% absolute improvement in Predicate A@1 under the incremental setting. Code and data are available at: https://github.com/Andy20178/SGFormer.

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

Retrieval-Augmented Foundation Models for Water Level Prediction in the Everglades

arXiv:2508.04888v2 Announce Type: replace Abstract: Accurate water level forecasting in the Everglades is essential for flood mitigation, drought management, water resource planning, and biodiversity conservation. While recent time-series foundation models have shown strong performance on generic tasks (represented in their pre-training), their effectiveness in domain-specific applications remains insufficiently understood. In this work, we curate a domain-specific dataset for water-level forecasting in the Everglades and observe that the performance of current state-of-the-art models remains limited. To address this gap, we leverage a retrieval-augmented mechanism that retrieves analogous multivariate hydrological episodes from an external archive of historical observations to enrich the input context of those pre-trained models. We study two retrieval strategies, statistical similarity-based retrieval and mutual information-based retrieval, and analyze how incorporating retrieved historical contexts affects predictive performance. Extensive experiments show that retrieval augmentation consistently improves long-horizon water level forecasts and yields disproportionately larger gains during extreme events, which is particularly critical for environmental decision-making. Our study provides empirical evidence that analog-based retrieval can benefit pretrained time-series foundation models in environmental science, offering practical insights into their strengths, limitations, and failure modes when applied to hydrological forecasting in the Everglades. Although evaluated in the Everglades, the proposed framework is general and can be applied to other hydrological systems given time series data. The code and data have been made publicly available at https://github.com/rahuul2992000/WaterRAF.