AI 速览：提出OopsieRecover框架，通过溯因推理从少量纠正示范中推断失败意图，并以稠密空间表示和条件代价引导VLA策略泛化纠正行为。
Vision-Language-Action (VLA) models can perform a wide range of robotic manipulation tasks, but often fail during deployment. Teaching them to recover from these failures requires substantial corrective data: existing approaches learn to recover from failures by directly associating corrective actions with the observations in which they were demonstrated. For instance, DAgger-like fine-tuning can require tens to hundreds of interactions, yet the resulting policy may still fail to generalize the learned corrections beyond the demonstrated scenarios. We hypothesize that semantics can capture the underlying reasons for failures and their corrections, enabling policies to learn from only a handful of corrective demonstrations and generalize those corrections to novel objects, and scene variations. We propose OopsieRecover, a framework that uses abduction to infer the intent behind a small set of corrective demonstrations and encodes the inferred knowledge as dense spatial representations and conditional costs. These costs capture when a behavior is undesirable and steer the policy toward more desirable action regions. We experiment across 6 tasks and variations in object poses, scenes, and related tasks, and find that only a handful of corrective demonstrations can enable a VLA to overcome failures across a broad range of novel situations, without requiring new corrective data for each variation. Please refer to https://robin-lab.cs.utexas.edu/oopsierecover for videos.
Radiologists interpret computed tomography (CT) scans by jointly determining what a finding represents and where it lies in three-dimensional (3D) anatomy. However, CT models have treated CT interpretation and 3D segmentation as separate tasks, leaving textual findings spatially unverifiable and overlooking potential synergy between clinical understanding and spatial perception. Here we present HoloCT, a generalist 3D foundation model unifying CT interpretation and voxel-level segmentation by jointly modelling whole-volume context and fine-grained anatomical detail, providing 3D masks as evidence. We curated HoloDATA, comprising 195,921 CT scans from 73 public datasets and six clinical cohorts, with 6.2 million instruction dialogues and 2.8 million CT-mask pairs. Across 346 tasks spanning segmentation, lesion screening, visual question answering, and report generation on 61 internal and 12 external datasets, HoloCT outperformed specialist and foundation models. In multicentre validation across 11,455 clinical CT scans from five hospitals, HoloCT outperformed competing models in lesion segmentation and report generation by 18.2 and 15.3 points. It also achieved 85.1% mean accuracy in screening thyroid tumours and colon lesions, exceeding all four participating radiologists in a retrospective reader study. Further analyses revealed bidirectional semantic-spatial synergy: language-guided segmentation delineated unseen anatomical structures and lesions as well as diffuse free-text findings, outperforming the leading approach by 22.1 points, whereas prompting the model to generate corresponding 3D masks alongside its textual outputs significantly improved interpretation performance. HoloCT shows that clinical understanding and spatial perception can be learned together within a single foundation model and offers a path towards more accurate and reliable CT analysis.
High-resolution video generation is expensive, as its cost grows rapidly with the number of spatiotemporal tokens. A practical alternative first generates a lower-resolution video and then applies a refiner, but conventional multi-step refinement introduces a second sampling bottleneck. We present SoL-Refiner, a one-step video refiner that transforms low-resolution model outputs into 4K videos with a single denoising step. Our three-stage recipe combines high-resolution continual training, reinforcement learning (RL) post-training, and a final one-step distillation. We introduce Refiner-Bench, a video refinement benchmark constructed from the outputs of different video generators, and use a shared-input protocol to compare refiners at approximately 2K output resolution. With LTX-2.3 Refiner as the base model, the one-step SoL-Refiner outperforms all external refiners at 2K on the VBench and UniPercept averages, while at $3840\!\times\!2176$ it improves both metrics over the three-step LTX-2.3 Refiner. With the complete acceleration stack, SoL-Refiner achieves an $8.91\times$ speedup in refinement latency over the same baseline in our 2K latency setting. With MiniMax H3 as the base model, SoL-Refiner produces 2K videos in three refinement steps and is preferred over the official MiniMax H3 2K API in 87% of user-study judgments.
Recent self-evolution frameworks for video understanding have demonstrated the potential for autonomous improvement from unlabeled videos. However, existing approaches often suffer from weakly controlled optimization and capability progression because they lack structured guidance for generating, evaluating, and reusing supervision throughout the iterative learning process. We present CurEvo, a curriculum-guided self-evolution framework that organizes this loop as a competence-aware progression. Curriculum-Guided Self-Evolution maintains a competence-aware curriculum state and cumulative training pool across iterations. Within this loop, Multi-Dimensional Question Generation supplies complementary supervision, from directly verifiable structural questions for perception to type-conditioned questions for recognition and understanding. Type-Adaptive Evaluation separately assesses question quality and answer reliability with criteria matched to each type's evidence; its scores determine which pairs enter the pool and how the next curriculum state is updated. Across seven backbones and four VideoQA benchmarks, CurEvo consistently improves benchmark accuracy and benchmark semantic score. For Qwen2.5-VL(7B), accuracy rises from 67.84% to 69.91% on ActivityNet-QA and from 76.94% to 79.61% on NExT-QA. CurEvo thus establishes self-generated supervision as a stable route to autonomous video understanding without task-specific human QA annotations during self-evolution.
Yulong Li, Rong Xia, Yuxuan Zhang, Jianxu Chen, Xiwei Liu, Haochen Xue, Maosheng Li, Yuhang Liu 等 15 位作者
We introduce DoAtlas-2, a foundation for self-evolving causal biomedical discovery that organizes knowledge around causal mechanisms and advances through external evidence from human populations. DoAtlas-2 integrates 771 research resources covering more than 720,000 participants in 48 countries, from longitudinal clinical phenotypes, medical imaging, and continuous physiological signals to eight molecular layers, together with an evidence network of approximately 4.7 million literature-derived records over 93,566 concepts and 149,383 candidate causal relations. DoAtlas-2 autonomously formulates research questions from evidence gaps and unresolved mechanisms, prespecifies their causal designs, and generates validated analyses. Supporting, challenging, and unresolved results continuously revise mechanistic interpretations, the causal evidence state, and the discovery frontier, so that DoAtlas-2 self-evolves within a closed loop of hypothesis generation, empirical testing, and renewed discovery. DoAtlas-2 has systematically evaluated 2,031 research questions. In the Human Phenotype Project (HPP), it formulated 4,014 candidate pathway questions across vascular, early-glycemic, and hepatic-metabolic systems, and screening of the first 1,079 yielded statistical support for 756. Representative studies identify blood pressure as a convergence node linking adiposity, hepatic, and lipid phenotypes to vascular outcomes, and show that an adiposity-inflammation-blood-pressure pathway is largely attenuated by joint adjustment for body mass index (BMI) and smoking. The discovered vascular network constitutes a completely interpretable predictive foundation, admitting exact attribution of every prediction and closed-form mediation effects. DoAtlas-2 thereby unifies causal mechanism discovery, population-evidence testing, and interpretable prediction within one continuously evolving foundation.
A text-to-SQL answer can run, return a plausible number and still compute something other than what the analyst meant. This can happen when the request leaves a calculation choice open, such as whether to count rows or distinct entities, and the model settles it without saying so. Standard benchmarks cannot detect this, because they score each question against a single reference interpretation. We introduce SpecBridge, which tests such a choice by varying it alone: it poses three otherwise identical requests, one leaving the choice open and one per interpretation, and scores each answer by running it. The databases are real, and the choices are ones that made models fail on existing benchmarks. On 204 hand-validated tests, four language models usually compute each interpretation when asked for it, yet for the open request return both in only 15-58% of the tests where they can, often without mentioning the choice. The gap lies in reporting, not in computing. Benchmarks even penalize the alternative: on BIRD questions where such a choice changes the answer, over two in five of the answers marked wrong are right under the other interpretation. Because the gap lies in reporting, a system can narrow it without a stronger model: it first decides whether a request leaves a choice open, then answers each interpretation in its own call. On our tests and on AMBROSIA, an existing benchmark of ambiguous questions, this approach raises the share of open requests that get every interpretation by 29-63 points over the single answer that production assistants give, without lowering precision on clear requests. It also needs fewer user turns than asking a clarifying question.
Scientific diagrams help researchers communicate complex methods and relationships, but creating them requires substantial effort in content organization, layout design, and visual refinement. Automated diagram generation can reduce this workload. Reference examples can guide this process by showing how scientific ideas are organized and presented visually. Diagrams from accepted papers provide a rich source of such examples, but their utility as a large-scale reference resource remains underexplored. We introduce DiagramBank, a diagram-text reference resource containing 57,100 scientific diagrams from 26,751 accepted AI/ML papers. Each diagram is paired with its caption, paper title, and abstract, with citing paragraphs included when available. We combine multi-stage filtering, automated checks, and human audits, with an estimated diagram-selection precision of 93.67%. Retrieval experiments show that citing paragraphs complement captions with information specific to individual diagrams. Under a fixed generation pipeline, replacing PaperBanana's curated reference pool with DiagramBank increases the overall PaperBanana score from 4.62 to 7.88 on PaperBananaBench. Retrieved references also outperform random references on BizGenEval scientific figures. Additional evaluations with different judges and renderers, together with a separate blind human evaluation, demonstrate the utility of DiagramBank for automated diagram generation. Code: https://github.com/csml-rpi/DiagramBank.
Conformal prediction (CP) provides marginal coverage guarantees, but marginal coverage does not determine where prediction errors occur on a graph. Queries from different graph neighborhoods can exhibit substantially different levels of uncertainty even when aggregate coverage remains near its target. We introduce GLocalCP, an online CP framework for inductive node classification and edge prediction that combines marginal and graph-local calibration with uncertainty-adaptive prediction sets. Specifically, GLocalCP puts forth a novel uncertainty-adaptive nonconformality score that relies on the mutual information obtained from Monte-Carlo (MC) dropout based prediction. Then, it augments a global online threshold with a kernel-based local correction that propagates coverage feedback among similar graph queries. On the theoretical horizon, we establish pathwise $O(N^{-1/2})$ marginal calibration error without exchangeability or stationarity and, for the exact-kernel formulation, an explicit $O(N^{-1/2})$ graph-localized calibration guarantee. Experiments on seven node-classification and seven edge-prediction benchmarks show that similar marginal coverage can conceal substantially different local reliability. Our proposed GLocalCP maintains coverage near the target while consistently reducing graph-local calibration error. Uncertainty scaling in GLocalCP further reduces prediction-set size across all evaluated node and edge datasets while leaving marginal coverage essentially unchanged.
Learn-then-Test (LTT) provides a general framework for certifying deployment policies against a prescribed risk target, but its standard formulation targets a fixed population risk estimated from calibration losses. In sequential deployment, losses may instead be dependent and nonstationary, so the risk of a fixed policy can evolve with the observed history. We introduce StreamLTT, which replaces this fixed population risk with average predictable risk, defined by conditioning each decision's loss on the information available before its outcome is revealed. Conditional centering turns empirical deviations from this target into a martingale sum, enabling both a test that adapts to predictable variance and a finite-sample Hoeffding alternative. Combined with family-wise error rate control, StreamLTT yields a simultaneously certified policy region that supports subsequent deployment selection. Across seven forecasting datasets and seven deployment risks, StreamLTT places the certified frontier substantially closer to the prescribed risk target than IID, robust-variance, block-bootstrap, and concentration-based alternatives, while empirical family-wise error closely follows the prescribed level. A finite-sample ablation further shows that the predictable-risk formulation drives the tighter risk frontier, while adaptation to predictable variance makes more effective use of the available family-wise error budget. StreamLTT therefore extends risk-certified policy selection to dependent deployment streams by addressing both what risk should be certified and how it should be tested.
Research taste, the judgment of what to read, which problems to pursue, and which methods to try, increasingly shapes the work of AI scientists, yet it remains far less understood than their ability to execute research. We introduce IdeaTaste, a benchmark that scores an LLM by how closely the research proposals it writes from a literature snapshot frozen at 2025 match the work the same communities published in 2026. Our experiments span six contemporary language models and four research areas from established to emerging fields, covering more than 90{,}000 proposals in total. LLM agents fall well short of human communities in both the literature they cite and the ideas they state, reaching about 26\% of the human level on citation matching and only 18\% on idea matching. Their populations can cite as diversely as a community and take up its tasks and methods, yet rarely arrive at the concrete approaches and research plans it pursues. Neither richer field context nor larger populations substantially close this gap. In blinded expert review, AI proposals trail the 2026 papers least in presentation and most in expected success, echoing their gap in concrete approaches, and those whose ideas no human paper shares fare worst, suggesting weaker directions rather than overlooked opportunities. These results indicate that the main bottleneck in AI research taste lies less in finding the directions a community works on than in turning them into the concrete approaches it goes on to pursue. We release IdeaTaste as an open, continuously updated leaderboard (https://cxcscmu.github.io/IdeaTaste) that benchmarks contemporary AI proposals against future publication cycles.
As general-purpose Large Language Models (LLMs), such as Gemini and DeepSeek, exhibit increasingly advanced reasoning capabilities, a critical question arises: how do their property prediction abilities compare to those of domain-specific pretrained materials models? In this study, we evaluate predictions for Ti, steels, high-entropy alloy (HEA) and Al, demonstrating that general-purpose LLMs surprisingly outperform SteelBERT, MatSciBERT, and SciBERT. To further enhance performance, we introduce a hybrid approach that utilizes numerical priors from a prior model, such as CatBoost or SteelBERT, to guide the general LLMs (Deepseek V3.1, Gemini-2.5-Flash). By inputting these initial estimates from a prior model as prompts, the LLM acts as a reasoning corrector, achieving accuracy superior to the prior model and "unguided mode". Furthermore, we investigate the impact of data strategy using a similarity-based metric, revealing that data quality dominates quantity; using merely 5% of high-similarity reference data of training set as input yields performance similar to the results of using 100% of training set. Additionally, the LLMs demonstrate excellent uncertainty calibration, where self-reported "high" confidence strongly correlates with minimal deviation from ground truth. Our findings suggest that synergizing general-purpose LLMs with prior model numerical baselines and high-quality data curation creates a cost-effective and accurate paradigm for materials discovery, reducing the need for expensive domain-specific pretraining.
Richard Yucheng He, Chen Xu, Yihang Liu, Tairan Chen, Baodong Cao
Most benchmarks for real-time voice systems focus on one aspect, such as turn-taking, instruction following, paralinguistic control, or conversational memory, each with its own data and set of systems. We present InteractionBench, a benchmark and evaluation harness that measures four dimensions on the same systems from the same recorded user audio: conversational timing, spoken-task accuracy, whisper and volume, and groundedness in conversations of more than twenty turns. The benchmark contains 221 items per system in thirteen scenarios. Before any system runs, we fix the expected floor-control actions for spoken tasks by rule and take the expected answers from scripted dialogues. We score whispering from the audio signal. We evaluate GPT-Live, GPT-Realtime, Gemini Live, Moshi, PersonaPlex, Freeze-Omni, and a turn-based cascade in 1,519 graded runs. Among the six full-duplex systems, timing and spoken-task accuracy are rank-correlated ($ρ= 0.83$), while the cascade has the highest spoken-task score (38/50) and the second-lowest timing score. No system corrects a user's grammar within the required window. Only Gemini Live whispers on most requests (8/10). Hosted systems pass 67-88% of groundedness probes, compared with 14-35% for open full-duplex models. Timing is measured where the client hears each system, so the timing of hosted systems includes network delay: for GPT-Live, whose server reports its own clock, removing a delay of about 1.5 s lowers its timing score from 86% to 56%. We also document where the protocol differs across systems, including the text instructions.
Gaussian process (GP) regression is a popular surrogate modeling tool for computer simulations in engineering and scientific domains. However, it often struggles with high computational costs and low prediction accuracy when the simulation involves too many input variables. For some simulation models, the outputs may only be significantly influenced by a small subset of the input variables, referred to as the ``active variables''. We propose an optimal kernel learning approach to identify these active variables, thereby overcoming GP model limitations and enhancing system understanding. Our method approximates the original GP model's covariance function through a convex combination of kernel functions, each utilizing low-dimensional subsets of input variables. Inspired by the Fedorov-Wynn algorithm from optimal design literature, we develop an optimal kernel learning algorithm to determine this approximation. We incorporate the effect heredity principle, a concept borrowed from the field of ``design and analysis of experiments'', to facilitate efficient computation. Through several examples, we demonstrate that the proposed method provides competitive prediction and variable selection, with favorable computational performance in the examples considered. It is an effective solution for interpreting the surrogate GP regression and simplifying the complex underlying system.
Action-conditioned world models are increasingly trained on data from randomized experiments, and whether their predictions respond to actions is usually checked by replacing or removing the actions and measuring the loss in accuracy. We show that such diagnostics report action effects where there are none: under autocorrelated designs, future actions stand in for history the model does not see, and a better model learns the stand-in better. When the design law is known, a conditional randomization test written as an e-value is valid in finite samples whatever the model, provided the resampler conditions on the full state the design used; resampling from a simpler law reproduces the false effect. We identify a second failure, specific to world models: when many sites share one action sequence (cells in a microscope field, pixels in a frame), the natural pooled score turns small model errors along the action contrast into overwhelming evidence against a true effect. In a pixel world, its median log e-value fell from $-17.3$ to $-6832$ as frames grew from 64 to 4096 pixels, while the oracle's rose. Tempering the score by a design effect estimated on training data changes the bet, not the validity. Our own preregistered analysis of optogenetic ERK experiments failed in exactly this way; a tempered statistic, selected on synthetic data and registered before the data were obtained, then rejected the null on a never-opened field ($E\approx 1.1\times10^{10}$). Every analysis is preregistered, fixed in a dated protocol before it was run, or labelled exploratory, and every reported result is generated from hashed result files.
AI 速览：该研究考察由大生成模型构成的多智能体系统在共享资源竞争、顺序交接与集体决策等场景中自发出现的群体性失败模式。
Multi-agent systems composed of large generative models are rapidly moving from laboratory prototypes to real-world deployments, where they jointly plan, negotiate, and allocate shared resources to solve complex tasks. While such systems promise unprecedented scalability and autonomy, their collective interaction also gives rise to failure modes that cannot be reduced to individual agents. Understanding these emergent risks is therefore critical. Here, we present a pioneer study of such emergent multi-agent risk in workflows that involve competition over shared resources (e.g., computing resources or market share), sequential handoff collaboration (where downstream agents see only predecessor outputs), collective decision aggregation, and others. Across these settings, we observe that such group behaviors arise frequently across repeated trials and a wide range of interaction conditions, rather than as rare or pathological cases. In particular, phenomena such as collusion-like coordination and conformity emerge with non-trivial frequency under realistic resource constraints, communication protocols, and role assignments, mirroring well-known pathologies in human societies despite no explicit instruction. Moreover, these risks cannot be prevented by existing agent-level safeguards alone. These findings expose the dark side of intelligent multi-agent systems: a social intelligence risk where agent collectives, despite no instruction to do so, spontaneously reproduce familiar failure patterns from human societies.
Multimodal large language models can recognize visual content yet struggle to reason reliably over it across multiple steps. In counting, a model may see all objects but lose track of which ones it has counted. We argue that such failures, commonly attributed to perception, reveal a reference gap: the failure to bind language to the relevant visual entities, regions, or locations. Motivated by the theory of visual routines, we test whether external references such as numbering can narrow this gap. We introduce ReGap-Bench, a paired benchmark of five visual reasoning tasks in which the scene, question, and answer are fixed and only the reference varies. Results show that task-matched references improve accuracy by up to 25\%, while mismatched references provide little benefit or even hurt performance. These findings identify the reference gap as a bottleneck distinct from visual perception and show that constructing task-appropriate visual references can serve as a training-free visual skill for improving multimodal reasoning.
Haochen Xue, Anastasia Godneva, Yulong Li, Qi Cao, Zhaodong Wu, Kaiyu Liang, Jionglong Su, Eran Segal 等 9 位作者
Existing gut microbiome foundation models typically use flat taxon representations, limiting their ability to exploit known taxonomic hierarchies. We introduce TaxoMFM, a microbiome foundation model that integrates taxonomic structure into self-supervised learning. TaxoMFM jointly learns abundance patterns across taxonomic ranks and relative proportions within groups through taxonomy-aware encoding and hierarchical compositional reconstruction. We further construct TaxoGut-HPP, comprising 100,806 gut microbiome profiles from 231 source cohorts and covering 40 classification and 8 regression tasks. TaxoMFM achieves competitive performance across disease, gut health, host trait, and liver ultrasound phenotype prediction. In cross-cohort zero-shot transfer, where target cohorts are excluded from both pretraining and downstream adaptation, TaxoMFM achieves AUROCs of 0.838, 0.736, and 0.722 for colorectal cancer, type 2 diabetes, and inflammatory bowel disease, respectively. Further analyses show that TaxoMFM recovers community changes across taxonomic scales, while preserving community information and producing consistent representations under database updates and species-vocabulary shifts. Our code is available at https://anonymous.4open.science/r/TaxoMFM-F16D/.
Rotation equivariance is an important structural prior for a broad range of vision tasks, particularly in domains such as remote sensing and medical imaging where objects have no canonical orientation, yet it remains notably absent from current Mamba-based vision architectures. Despite the success of Mamba in natural language processing and its growing adoption in computer vision, existing visual Mamba models fail to account for rotational symmetry in their design. This omission renders them sensitive to image rotations, thereby constraining their robustness and cross-task generalization. To address this limitation, we incorporate rotation symmetry, an important geometric prior in many visual domains, into Mamba-based architectures. Specifically, we introduce EQ-VMamba, the first rotation equivariant visual Mamba architecture for vision tasks. The core components of EQ-VMamba include a carefully designed rotation equivariant cross-scan strategy and group Mamba blocks. Moreover, we provide a rigorous theoretical analysis of the intrinsic equivariance error, demonstrating that the proposed architecture enforces end-to-end rotation equivariance under the p4 rotation group. Extensive experiments across multiple benchmarks---including high-level image classification, mid-level semantic segmentation, and low-level image super-resolution---demonstrate that EQ-VMamba consistently improves rotation robustness and achieves superior or competitive performance compared to non-equivariant baselines on most benchmarks, while requiring 40-70% fewer parameters. These results indicate that embedding rotation equivariance not only effectively bolsters the robustness of visual Mamba models against rotation transformations, but also improves accuracy on data exhibiting rotational symmetry, with substantially improved parameter efficiency. Code is available at https://github.com/zhongchenzhao/EQ-VMamba.
Rubric-based reinforcement learning (RL) offers interpretable, fine-grained supervision for aligning large language models (LLMs) on open-ended generation tasks. However, static rubric criteria may become saturated or remain beyond the policy's exploration, rendering the reward signals insufficiently discriminative. While recent efforts seek to guide sampling exploration or evolve rubrics, they fail to explicitly factor in how rubric difficulty should track the evolving policy's capabilities, resulting in a persistent mismatch between the optimization objective and the policy's current capability frontier. In this paper, we propose \textbf{Co}-Evolving Policy and Rubric via \textbf{C}apability \textbf{A}wareness (\textbf{CoCA}), a closed-loop framework that jointly adapts the rubric and policy through criterion-level diagnosis. Specifically, CoCA contrasts query-only and rubric-guided rollouts to characterize each criterion relative to the current policy, recalibrates criteria with mismatched difficulty, and identifies guidance-revealed capability gaps for frontier-based learning. Based on this insight, CoCA either expands rubric-conditioned frontiers through guided optimization or transfers qualified gains through on-policy self-distillation. As the policy updates, CoCA improves the capability diagnosis and enables more reliable distillation, thereby completing the co-evolution loop. Extensive experiments across diverse benchmarks and LLMs demonstrate that CoCA consistently improves performance over rubric-based RL methods. Moreover, CoCA promotes the internalization of guidance-revealed capabilities into query-only behavior while mitigating capability degradation.
AI compute is becoming large and fast-changing electrical loads. Its impacts on data center infrastructure and power grids depend not only on total energy consumption, but also on how their power draw moves over sub-second to minute timescales, including peaks, ramps, and synchronized fluctuations across accelerators. Existing energy benchmarks for large language models (LLMs) report aggregate quantities such as joules per token and rarely release the underlying time-series measurements, leaving these power dynamics poorly characterized and difficult to model. We introduce LLM-PowerTrace, an open dataset of time-resolved, per-GPU power traces for LLM fine-tuning and inference, sampled every 100\,ms on varying NVIDIA platforms paired with complete workload and system metadata. For fine-tuning, LLM-PowerTrace profiles open model families spanning a wide range of model sizes and fine-tuning methods, together with controlled sweeps over batch size, sequence length, precision, parallelism, and checkpointing. For inference, it covers variations in concurrency, GPU frequency, KV-cache and weight quantization, prompt and output lengths, and request-arrival patterns. LLM-PowerTrace profiles a set of open model families across a wide range of model sizes, and controlled sweeps over batch size, sequence length, precision, parallelism, and checkpointing for fine-tuning, and it sweeps concurrency, GPU frequency, KV cache and weight quantization, prompt and output lengths, and request arrival patterns for inference. Built upon power traces, we define power-grid-facing metrics as well as AI-compute power\&energy scaling law for extrapolating beyond the measured settings. Together, LLM-PowerTrace provides a dataset foundation for modeling, forecasting, and managing the fast-changing electrical behavior of AI compute.
We present the first optimistic concurrent B-skiplist, built on optimistic lock coupling (OLC). OLC replaces the multi-cache-line RW lock with a compact, single-cache-line lock: readers validate a version number without writing to shared memory, avoiding cache invalidations and retrying only on a detected conflict. Although OLC was thought poorly suited to skiplists, where a single per-element lock may span structurally unrelated levels, the B-skiplist's blocked layout places each level (and associated lock) in its own node. We exploit the B-skiplist's upfront knowledge of promotion height with a hybrid top-down scheme that traverses upper levels optimistically and locks only from the promotion level down, so most inserts never dirty a cache line. On 128 threads, the proposed optimistic B-skiplist achieves $1.2-10\times$ higher throughput over state-of-the-art concurrent skiplist implementations. Furthermore, the optimistic B-skiplist achieves $0.8-4.3\times$ the throughput compared to high-performance optimistic B+-tree implementations based on RW locks and OLC. Finally, the optimistic B-skiplist achieves up to $13\times$ lower latency than concurrent skiplists and tree-based indices, and is within $1.6\times$ of the latency of the OLC-based B+-tree.
As Large Language Models (LLMs) are increasingly integrated into software development workflows, they also become prime targets for adversarial attacks. Among these, backdoor attacks are a significant threat, allowing attackers to manipulate model outputs through hidden triggers embedded in training data. Detecting such backdoors remains a challenge, and one promising approach is the use of Spectral Signature defense methods that identify poisoned data by analyzing feature representations through eigenvectors. While some prior works have explored Spectral Signatures for backdoor detection in neural networks, recent studies suggest that these methods may not be optimally effective for code models. In this paper, we revisit the applicability of Spectral Signature-based defenses in the context of backdoor attacks on code models. We systematically evaluate their effectiveness under various attack scenarios and defense configurations, analyzing their strengths and limitations. We found that the widely used setting of Spectral Signature in code backdoor detection is often suboptimal. Hence, we explored the impact of different settings of the key factors. We discovered a new proxy metric that can more accurately estimate the actual performance of Spectral Signature without model retraining after the defense. Our findings contribute to a deeper understanding of the configurations of Spectral Signature defenses in securing code models and offer insights into potential improvements for future research.
Visual counting benchmarks report whether vision-language models produce the correct number, but not whether errors arise from perception or reasoning. We introduce DecoyCount, a benchmark built from 1,660 real-world images with designed decoys across seven mechanisms, along with paired prompts: a neutral question, a matched alignment cue, and a novel multi-hop, multi-image question that situates the same counting problem in a realistic scenario. Using this design, we find that counting failures are induced partly by visual misalignment: models can execute the intended counting process when cue-guided but fail to select it unaided, and we term this difference in accuracy the reasoning-alignment gap. Despite being highly capable on general reasoning, frontier models remain below 50\% accuracy on our benchmark, while alignment gaps range from 11.8-22.1 percentage points; strikingly, we observe that more capable models, greater thinking budgets, and agentic processes often induce larger gaps. Because humans may bring context that models lack, we test whether textual underspecification contributes to misalignment using a scenario-based multi-hop condition without explicit hinting, but this added context does not improve aggregate performance. Our analysis concludes that stronger general reasoning does not by itself generalize to aligned visual reasoning, leaving misalignment with human visual expectations as a substantial component of the remaining deficit.
Motasem Alkayid, Zhengsen Xu, Naser El-Sheimy, Lincoln Linlin Xu
Tree species classification from Moderate Resolution Imaging Spectroradiometer (MODIS) time series is critical for environmental monitoring, yet challenging due to subtle inter-species signatures, strong spatial-spectral-temporal coupling, and the difficulty of modeling large-scale geographical correlations. This letter presents a Graph-regulated Disentangled Sparse Mamba model (Geo-GDS-Mamba) with three contributions. First, a mini-batch graph-regulated stream models geographic correlations among samples via a Radial-Basis-Function kernel over their physical coordinates, independent of spectral content, and fuses this context through a graph convolutional network without full-batch memory costs. Second, a disentangling Mamba architecture separates the spatial, spectral, and temporal information coupled within each sample into three independent processing streams. Third, sparse token modules adaptively retain the top 30\% most informative tokens per branch via learned attention or similarity scores, mitigating Mamba's correlation-decay limitation while reducing computational cost. Experiments on annual MOD13Q1 data across two Canadian provinces show Geo-GDS-Mamba achieves 95.35\% OA (94.73\% Kappa) in Alberta, outperforming eleven state-of-the-art baselines, and generalizes without fine-tuning to unseen Saskatchewan data at 87.63\% OA, using only 1.74M parameters and 0.18 GFLOPs. Ablation studies confirm each component's contribution, led by a 4.15-point OA gain from the geographic graph stream alone.
Can AI agents autonomously train useful language models and improve them using feedback from users? We introduce RSI Arena, a two-stage study of autonomous post-training as a step toward recursive self-improvement. Ten frontier models act as training agents, each starting from NVIDIA Nemotron 3.5 Lightning 30B-A3B Base. During a six-day first stage, each agent has 1,000 GPU-hours and $300 in API credit on a shared 64-GPU cluster, independently selecting data, implementing training procedures, and choosing checkpoints. We evaluate their models on 20 benchmarks and through blind human evaluations. All ten agents produce working assistants, with the strongest model, trained by GPT-6 Astra, closing a median of 76% of the base-to-Instruct gap across 12 capability benchmarks. Its human-preference rating is also close to NVIDIA Instruct after statistical adjustment for answer length and formatting. In Stage 2, each agent receives the comparison records and votes involving its Stage 1 model and is given three additional days for further training. Using NVIDIA Instruct as a common reference across the two evaluation rounds, we observe an improvement from Stage 1 to Stage 2 in the style-adjusted relative rating of eight of the ten models. We show that benchmark gains were modest, with no clear link to preference gains, while some agents encouraged longer responses and one model showed increased harmful compliance on StrongREJECT. RSI Arena provides an auditable setting for studying how human feedback can guide autonomous model development. It takes a step toward recursive self-improvement with human feedback (RSIHF), in which people help define what improvement means for the AI systems they use.
Rui Yin Zhejiang University, Tianxu Han, Naen Xu Zhejiang University, Changjiang Li Palo Alto Networks, Ping He Zhejiang University, Chunyi Zhou, Jun Wang OPPO Research Institute, Zhihui Fu Guangdong OPPO Mobile Telecommunications Corp.,Ltd. 等 11 位作者
Safety-aligned large language models (LLMs) are increasingly deployed in real-world pipelines, yet this deployment also enlarges the supply-chain attack surface: adversaries can distribute backdoored checkpoints that behave normally under standard evaluation but jailbreak when a hidden trigger is present. Recent post-hoc weight-editing methods offer an efficient approach to injecting such backdoors by directly modifying model weights to map a trigger to an attacker-specified response. However, existing methods typically optimize a token-level mapping that forces an affirmative prefix (e.g., “Sure”), which does not guarantee sustained harmful output—the model may begin with apparent agreement yet revert to safety-aligned refusal within a few decoding steps. We address this reliability gap by shifting the backdoor objective from surface tokens to internal representations. We extract a steering vector that captures the difference between compliant and refusal behaviors, and compile it into a persistent weight modification that activates only when the trigger is present. To preserve stealthiness and benign utility, we impose a null-space constraint so that the injected edit remains dormant on clean inputs. The method is efficient, requiring only a small set of examples and admitting a closed-form solution. Across multiple safety-aligned LLMs and jailbreak benchmarks, our method achieves high triggered attack success while maintaining non-triggered safety and general utility.
尚无 AI 解读，展示原文摘要
cs.CL
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表
Chen Xu, Yu Ji Beijing Institute of Technology, Zhenyu Lyu, Yang Yi, Yizhe Yang, Luyao Ji, Chaoyi Chen, Xianyang Wang 等 15 位作者
The critical therapist shortage demands scalable training solutions. Standardized Patients, the gold standard, are scarce and costly. Current LLM-based approaches focus on patient simulation for conversational realism but lack pedagogical rigor as Virtual Standardized Patients, lacking faithful reactions to clinical errors and explainable feedback. To bridge this gap, we propose PUPPET, the first neural-symbolic Virtual Standardized Patient governed by an OBSERVE-THINK-BEHAVE architecture. PUPPET externalizes LLM reasoning into a symbolic system where experts implant causal associations between intervention logic (propositional logic) and patient mental states (state machine). This allows PUPPET to behave coherently with controllable and explainable psychological dynamics: intervention logic (OBSERVE) → state transition (THINK) → response (BEHAVE). Our PUPPET-TRAINER further leverages this chain to educate trainees about intervention consequences, standardizing and scaling mental health training. Experiments across three clinical scenarios confirm that PUPPET outperforms baselines in clinical faithfulness and pedagogical value.
尚无 AI 解读，展示原文摘要
cs.CL
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表
Yuetai Li, Zhangchen Xu, Fengqing Jiang Apple and University of Washington, Bhaskar Ramasubramanian Western Washington University, Luyao Niu University of Washington, Bill Yuchen Lin xAI and University of Washington, Xiang Yue Carnegie Mellon University, Radha Poovendran University of Washington, Seattle
Fine-tuning large language models (LLMs) is intended to improve their reasoning capabilities, yet we uncover a counterintuitive effect: models often forget how to solve problems they previously answered correctly during training. We term this phenomenon Temporal Forgetting and show that it is widespread across model sizes, fine-tuning methods (both Reinforcement Learning and Supervised Fine-Tuning), and multiple reasoning benchmarks. Our analysis reveals on average more than 20% of final errors were once solved correctly at an earlier checkpoint. Inspired by the phenomenon of Temporal Forgetting, we proposed Temporal Sampling, a simple decoding strategy that draws outputs from multiple checkpoints along the training trajectory. This approach recovers forgotten solutions and leads to significant improvements in reasoning performance than final-ckpt-sampling only, gains from 4 to 19 points in Pass@k and consistent gains for majority-voting and Best-of-N across several benchmarks. Temporal sampling also outperforms strong baselines such as model merging. By leveraging the temporal diversity inherent in training, Temporal Sampling offers a practical, compute-efficient way to surface hidden reasoning ability and rethink how we evaluate LLMs.
尚无 AI 解读，展示原文摘要
cs.CL
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表
Jayeol Chun Brandeis University, Nianwen Xue Brandeis University
Modal dependency parsing-the task of identifying a semantic graph that represents who is responsible for an event-centered claim and with what degree of certainty-relies on recognizing source-introducing cues and correctly linking them to their associated content. However, prior work has largely focused on identifying sources only, treating cue expressions and their modal coverage as auxiliary signals. In this work, we propose a structured prediction framework that leverages large language models (LLMs) to explicitly identify source-cue pairs as well as their respective scope, which together define the modal contexts governing downstream source attribution for events. By concentrating learning at the source-cue level and constraining event-level decisions to a small, scope-defined candidate set, our top-down approach enables more efficient inference in long, event-rich documents. Experiments show this approach surpasses prior state-of-the-art results by 3 and 4% for English and Chinese datasets, respectively.
尚无 AI 解读，展示原文摘要
cs.CL
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表
Large Language Models (LLMs) are increasingly deployed in socially sensitive domains, yet their unpredictable behaviors, ranging from misaligned intent to inconsistent personality, pose significant risks. We introduce SteerEval, a hierarchical benchmark for evaluating LLM controllability across three domains: language features, sentiment, and personality. Each domain is structured into three specification levels: L1 (what to express), L2 (how to express), and L3 (how to instantiate), connecting high-level behavioral intent to concrete textual output. Using SteerBench, we systematically evaluate contemporary steering methods, revealing that control often degrades at finer-grained levels. Our benchmark offers a principled and interpretable framework for safe and controllable LLM behavior, serving as a foundation for future research.
尚无 AI 解读，展示原文摘要
cs.CL
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表