We present a spatial-memory system that enables optical see-through AR glasses to recall earlier physical encounters. The dual-camera prototype combines hierarchical visual acquisition with source-linked storage and staged retrieval. Quality screen-ing, global and regional appearance novelty, and temporal support select monochrome observations for detailed RGB cap-ture. Semantic observations retain links to their source images, allowing retrieval to answer indexed historical questions direct-ly or inspect visual details omitted from scene descriptions. The architecture separates capture acceptance, interpretation out-come, and support for a particular question. This distinction motivates two core control mechanisms: bounded reassessment of familiar views when earlier processing leaves an evidence gap, and retention guided by source dependencies. In a 131.35 s egocentric replay, the implemented acquisition policy selected six RGB views from 648 monitoring observations, while feature extraction ran on 97.7% of observations. These results demon-strate selective RGB acquisition and identify frequent local inference as a remaining cost. Exposure duration and cross-view timing further constrain the evidence preserved from se-lected encounters. The system provides a practical acquisition-to-recall architecture for persistent AR assistance, with explicit boundaries between captured evidence, indexed interpretations, and supported historical claims.
Finite-precision arithmetic unavoidably introduces numerical approximation errors. Numerical computations may use insufficient precision or an improper formulation, which leads to numerical instability. In this paper, we introduce a unified software tool for stochastic numerical validation of deep-learning operators. The tool follows CESTAC on supported exposed operations and uses an operator-level data-perturbation approximation for GEMM-like kernels. Our developed software not only enables numerical validation with a single computation pass but also detects the sources of numerical instability and provides numerical stability monitoring during deep learning training and inference. We verified its effectiveness on the detection of polluted operators with injected numerical instabilities across various tasks. We believe that our developed method and tools provide valuable insights into developing numerically stable computing kernels, which are particularly critical for numerically stable and efficient deep learning training and inference.
Yeonsu Kwon, Jiho Kim, Junseong Choi, Paloma Rabaey, Minseo Kim, Sujeong Im, Jeewon Yang, Jun-Min Lee 等 13 位作者
Data consistency between unstructured clinical notes and structured tables in Electronic Health Records (EHRs) is essential for patient safety and clinical decision-making. However, existing work on note-table consistency verification mainly relies on surface-level matching of numeric values or simple events. Such approaches fail to capture the reasoning underlying real-world EHR documentation, including clinical interpretation, event relations, and temporal changes. To address this gap, we introduce EHR-ReasonCon, a reasoning-intensive benchmark for note-table consistency verification. Built on MIMIC-III with expert-guided annotations, it comprises 8,042 entities derived from clinical notes and provides high-quality ground-truth labels. The annotation protocol is supported by specialized table-exploration tools to ensure systematic evidence retrieval and reliable consistency assessment. We also propose EHR-INSPECTOR, an LLM-based framework that segments notes, extracts anchor entities and temporal references, and uses table-exploration tools to verify consistency against structured tables. Evaluated using expert-validated LLM-as-a-judge metrics under harsh and lenient criteria, EHR-INSPECTOR achieves state-of-the-art performance across multiple model backbones. Analyses further demonstrate the effectiveness of its components and highlight differences from human verification. The code and dataset are available at https://github.com/dustn1259/EHR-ReasonCon.
Alessandro Zirilli, Davide Marincione, Evgenios M. Kornaropoulos, Giuseppe Ateniese, Emanuele Rodolà
Large Language Models (LLMs) have made it exceptionally convenient for users to send their private data to model providers, which process them in plaintext. Fully Homomorphic Encryption (FHE) promises to secure remote inference by computing directly on encrypted data, yet it has remained prohibitively slow. The effort to make it practical remains fragmented, with most existing frameworks being closed-source, which keeps progress from compounding. We present Perseus, a free and open Python software library that supports the full stack of encrypted model inference, from the cryptographic backend and data packing, to encrypted nonlinearities and end-to-end model execution. Perseus ships with built-in novel optimizations, one at each level of the stack: in-kernel regeneration of key-switching keys at the (i) backend level, a complex variant of Cachemir packing at the (ii) data level, and a state-of-the-art bootstrap placer that is both packing- and magnitude-aware at the (iii) model level. Perseus achieves state-of-the-art latency across multiple models and modalities, within $3\%$ of the plaintext models' accuracy and perplexity. Remarkably, it is $3.96\times$ faster at single-token decoding than any other implementation on Llama-3-8B. We initiate, for the first time, a public speedrun on encrypted GPT-2, so that progress can compound.
Oliver Niggemann, Jonas Ehrhardt, Alexander Diedrich
Not every component in a technical system can be equipped with sensors. Hence, when a fault occurs, available sensors often detect it only after its effects have propagated. This makes the isolation of faults difficult. We approach this problem by introducing virtual sensors. For each combination of component and fault type, we let external knowledge describe the fault-sensitive variables. We introduce a novel neural observer that is only trained on normal data. For fault isolation, it uses available sensor measurements where it is possible and reconstructs missing variables as virtual sensors to predict normal behavior. It detects deviations between current values and nominal predictions and computes local symptoms, which are then compared by onset time. The result is a set of root causes describing the faulty behavior. We use Gödel logic to encode physical knowledge and guide training. We evaluate our approach on synthetic networks, the Tennessee Eastman Process, and datasets of two modular production plants.
Xuwei Tan, Jingchun Huang, Ming Cai, Shuangquan Lyu, Tiange Wang, Yinzhi Wang, Ronghua Shi, Jingwei Song 等 15 位作者
Multi-expert agent marketplaces -- where creators publish domain-specific AI experts for end users to consult -- are an emerging deployment pattern for language technology. A published expert can receive many requests yet little evidence about its creator, who remains active elsewhere: the cold-expert problem. We frame continuous cognitive distillation as converting attributed creator observations into revisable decision rules, communication patterns, and domain procedures. CogniLife proposes a cross-expert attribution design for this problem, with Creator Profiles for transferable creator patterns and Expert Journals for listing-specific knowledge; this design is not yet implemented or evaluated. A deployed platform partially realizes the surrounding lifecycle through creator-authorized authoring, capability-owned proposal filtering, and reversible consolidation of eligible owner-scoped memory. On ProfyBench-v1, an 85-task offline benchmark built from replay seeds and controlled internal tests, the Profy Full configuration as a whole achieves a mean overall score of 4.65/5 from the primary LLM judge, compared with 4.01 for Raw LLM and 4.28 for Expert Prompt, with the largest gains in completeness, specificity, and missing-information tasks. These configuration-level scores do not test cold-expert mitigation, cross-expert transfer, or longitudinal distillation.
Tianxing Zhu, Huacheng Lu, Ming Liu, Malte Hillebrand
We propose goal-driven self-recursive optical research in which agents use physical evidence to revise the hardware, software, models, and protocols used to pursue a scientific goal. The framework can be instantiated with different stimulus systems and measurement devices. We formalize its updates and derive a relation between physical attainability, command access, and model approximation error. We demonstrate the concept in an optical see-through color-matching case study using an iPad, RayNeo X3 glasses, and a TOP 300 probe coupled to a CAS140D spectroradiometer, both from Instrument Systems. A revised glasses application adds USB-controlled RGB holds and frame acknowledgements, keeping model-generated colors displayed during spectral acquisition. The interface passes 12 live control checks. Spectral feedback reduces one target color difference from 19.756 to 1.211 ΔE_76^*; all five targets improve in a separate comparison. These measurements demonstrate the concept in one configuration. The resulting software and protocols remain available for subsequent investigations.
Michel Barlaud, Clement Genninasca, Maxime Zanella, Benoit Macq, Benoit Legat
In this paper, we introduce the Cone Alignment Index (CAI), a scale-invariant sparsity measure associated with a Lorentz-cone geometry. We provide a probabilistic analysis of the distribution of the CAI score, including a general theorem and explicit corollaries for Gaussian and uniform distributions. As a consequence, we show that, in the high-sparsity regime, the probability that the samples lie within the convex cone tends to one. Building upon this formulation, we derive two complementary projections: the \emph{Parametric Linear Conic Projection} (PLCP), an efficient closed-form approximation, and the \emph{Parametric Euclidean Conic Projection} (PECP), the exact Euclidean projection onto the cone boundary. Our analysis establishes several fundamental theoretical properties. We prove that PLCP and PECP are collinear, differ only by a positive radial scaling factor, and induce exactly the same sparsity threshold, thereby preserving an identical active support. Furthermore, we derive a support-identification theorem showing that, for any prescribed sparsity level $l<n$, the number of coefficients discarded at each iteration is theoretically controlled. Building on this theoretical foundation, we develop a bilevel PLCP framework for structured matrix sparsification. The proposed method naturally generates hardware-friendly structured sparsity patterns, including row-wise, column-wise, and diagonal masks, making it particularly well suited for accelerating Transformer attention mechanisms.
Multi-view spatial reasoning requires vision-language models to compare visual evidence across images, align object correspondences, and infer spatial relations over long visual contexts, a setting where chain-of-thought reasoning tends to grow verbose without becoming more accurate. Reinforcement learning with verifiable rewards is a natural fit for this task, but standard GRPO reward relies on sparse outcome-level feedback and gives no signal about where a reasoning trajectory goes wrong, nor any control over its length. We propose LenGuard-GPC, a dense reward framework that addresses both problems together. For each sampled trajectory, it compares the token-wise predictive distributions under a standard prompt and a guided prompt, and uses the resulting token-sum KL divergence as a dense reward signal. Since this KL penalty accumulates over tokens and would otherwise reward shorter responses regardless of their quality, we introduce a staged length bonus that keeps reasoning length within a controlled range without simply encouraging brevity. On six multi-view spatial reasoning benchmarks, LenGuard-GPC improves accuracy over vanilla GRPO while reducing average response length.
Anna Gogleva, Thomas Charman, Martin Balla, Karl Gemayel
Scientific agents can find interventions that work without understanding why they do. We study this failure mode in controlled synthetic regulatory systems, where agents must experimentally identify a hidden fault and rescue the resulting phenotype. Across seven LLMs, every model restores the healthy phenotypes more often than it identifies the fault. Claude Opus 4.6 and Sonnet 4.6 rescue every organism while diagnosing only 55% and 35% of them. This failure is not simply due to insufficient exploration or unavailable evidence. In most cases, the cause is experimentally identifiable, and agents encounter evidence that distinguishes competing causes in 91% of relevant sessions, and still frequently misdiagnose the system. Targeted probes reveal where experimental reasoning breaks down. Models reliably interpret decisive evidence and revise hypotheses when the causal comparison is explicit, and can locate that evidence within a full experimental record. However, they struggle to predict the outcome of the experiment under each competing cause, a necessary step in using the observed result to distinguish between them. Agents can therefore gather the right evidence and still infer the wrong cause. As AI systems take on a growing role in synthesising scientific evidence, these findings highlight the need to evaluate how they use evidence to reach causal conclusions, alongside whether their interventions succeed.
Xiaoke Yan, Jiaying Lei, Danqing Shi, Zichen Wang, Nan Cao
Recent advances in large language models (LLMs) have enabled coding agents to autonomously perform complex software development tasks, yet their long-running execution remains difficult for users to inspect, understand, and steer. We present Vibetrace, an interactive trace visualization and intervention system for agentic programming built upon OpenCode. Vibetrace introduces a human-agent collaboration paradigm of Observe, Understand, Intervene, and Evolve. It (1) visualizes agent workflows as interactive execution traces for inspecting actions, intermediate states, and failures; (2) enables users to fork execution from prior steps and provide targeted corrections without restarting validated progress; and (3) distills traces and human feedback into reusable skills for continual agent evolution. On SWE-bench Verified, distilled skills reduce input tokens by 23.0% and cost by 15.5%, while improving pass rate from 65.0% to 68.0%. User studies further demonstrate improved accuracy and efficiency over conventional log-based interaction.
Rong Shan, Aofan Yu, Bo Chen, Kuo Cai, Qiang Luo, Ruiming Tang, Han Li, Weiwen Liu 等 10 位作者
Recommender systems (RecSys) are increasingly emphasizing scaling, leveraging larger architectures and more interaction data to improve personalization. Yet, despite the optimizer's pivotal role in training, modern RecSys pipelines almost universally default to Adam/AdamW, with limited scrutiny of whether these choices are truly optimal for recommendation. In this work, we revisit optimizer design for scalable recommendation and introduce MuonRec, the first framework that brings the recently proposed Muon optimizer to RecSys training. Muon performs orthogonalized momentum updates for 2D weight matrices via Newton-Schulz iteration, promoting diverse update directions and improving optimization efficiency. We develop an open-source training recipe for recommendation models and evaluate it across both traditional sequential recommenders and modern generative recommenders. Extensive experiments demonstrate that MuonRec reduces converged training steps by an average of 32.4\% while simultaneously improving final ranking quality. Specifically, MuonRec yields consistent relative gains in NDCG@10, averaging 12.6\% across all settings, with particularly pronounced improvements in generative recommendation models. These results consistently outperform strong Adam/AdamW baselines, positioning Muon as a promising new optimizer standard for RecSys training. Our code is available.
Robin Na, Kehang Zhu, Logan Cross, Michiel A. Bakker
Large language models (LLMs) are increasingly used to simulate human behavior, but which descriptions of people improve prediction when experimental designs change remains unclear. We develop a framework that draws candidate behavioral features from the literature and constructs profiles from earlier human experiments. An LLM-based profile generator assigns feature values from observed actions, messages, and surveys; a separate LLM simulates behavior using these profiles. By comparing simulated and observed outcomes in the training data, we select profile contents through forward feature selection, adding features while they improve prediction. We then evaluate the selected contents without modification against outcomes from new participants under new designs. Across 616 public-goods games, 144 multi-party bargaining games, and 1,480 generalized 11--20 money-request designs, leading one-feature profiles predict human outcome distributions better than no profile or general-purpose personas. Profiles with more features provide no clear additional benefit, and features' relative predictive value largely persists across learning and validation. Using profiles from one game family to simulate another produces uneven gains and can make predictions less accurate than simulations without profiles. The framework makes the choice of behavioral information a testable part of simulation design, helping researchers assess which descriptions remain useful as experimental conditions change.
Low-rank adaptation (LoRA) has attracted significant attention in the parameter-efficient fine-tuning (PEFT) landscape. However, its design suffers from two fundamental limitations. First, its invariance manifold induces flat directions in the loss landscape, creating an optimization bottleneck for adaptive optimizers. Second, its effective rank scales only linearly with the parameter budget. In this paper, we propose Scalable Morpho Adaptation (SaMA), a PEFT method grounded in a principled generalization of the Kronecker product. By relaxing the shared-block constraint in the perfect-shuffle factorization, SaMA spans from low to full rank and achieves a practical rank that scales quadratically with the parameter budget, up to full-rank. Furthermore, we show that its invariance manifold collapses to a diagonal group, making it structurally more optimizer-friendly. Empirically, SaMA consistently outperforms strong PEFT baselines on commonsense and arithmetic reasoning benchmarks while being more parameter-efficient.
Exchange-correlation functionals are usually fitted with energy-only losses, so better energies often bring worse densities. We introduce a workflow that predicts in closed form, from one Fock build on inverse Kohn-Sham orbitals of any accurate density, a functional's self-consistent density, and fits it to that density with no self-consistent-field solution. Fitted convexly, a CF22D-form hybrid beats CF22D and MN15 on unseen energies, dipole moments, and unfitted densities (6.8- and 22-fold lower density error). A neural meta-GGA fitted to 6 energies and 11 densities beats SCAN in both.
While conversational recommendation systems provide a more natural user interface, they face the challenge of progressively discovering a user's needs before they can recommend an item that satisfies them. To evaluate the ability of AI agents at this elicitation process, we introduce $τ$-Rec-Books, a benchmark that holds each user request fixed while varying only when the user reveals their requirements: all at the start (volunteer), one at the start and the rest when asked (mixed), or some never (hidden). A hidden requirement can be found only by proposing books and inferring it from the user's accept or reject verdicts. We generate 450 tasks over a catalog of 20,000 books from Wikidata, develop a rule-based simulated user to answer the agent, and score recommendations against catalog metadata without an LLM judge. Across four agents and two search interfaces, success rates fall for every agent and interface, on average from 0.975 (volunteer) to 0.756 (mixed) and 0.283 (hidden). Mixed tasks fail mainly because agents do not ask, and when told to ask, every agent reaches at least 0.931. On hidden tasks, most failures come from agents recommending a book they never proposed to the user. When told to propose first, agents only match a simple baseline that proposes popular books, and stay well below one that uses rejection to rule out possible requirements. We release the catalog pipeline, task generator, simulated user, validators, reference agents, and all agent traces at https://github.com/nbharaths/tau-rec/tree/main/data/tracks/books-wikidata-v2.
He Wang, Yeye Liu, Ruijie Li, Xuemei Jiang, Junyu Wu, Lining Liu Zhuoqi Man, Tao Li, Yanjie Song
Agile Earth observation satellites must coordinate imaging tasks within limited observation windows, while accounting for attitude maneuvering and onboard energy constraints. Competition among observation opportunities makes scheduling performance sensitive to the heuristic rules used for task selection. This letter proposes a large language model (LLM)-assisted adaptive heuristic search method for agile Earth observation satellite scheduling. The LLM generates an initial heuristic rule according to the characteristics of the scheduling instance. During the subsequent ant colony search, search performance and scheduling information are fed back to the LLM to refine the heuristic rule. The proposed modifications are evaluated through paired solution constructions before being incorporated into the search process. A unified scheduling procedure determines task selection, satellite assignment, and observation timing under the operational constraints. Experiments on six scheduling scenarios involving 50--200 imaging tasks demonstrate that the proposed method achieves 4.3\%--9.6\% higher mean observation benefit than the strongest baseline.
Using large language models to interpret experimental evidence, design discriminating experiments and formulate well-supported mechanistic explanations remains a central challenge in materials research. Here we establish PSM-Bench, an 800-question benchmark developed by perovskite researchers to evaluate scientific reasoning in perovskite photovoltaics. We also introduce OpenPerov, an expert-informed language-model framework that learns from expert-calibrated evidence records reconstructed from 4,035 training articles; these records link experimental strategies, controls and observations to mechanistic interpretations. Structured-evidence adaptation outperformed conventional question-answer training, and OpenPerov Flash ranked first among 13 closed-book systems. OpenPerov Pro raised Scientific Mastery further, from 87.6 to 89.1, by revising selected passages with retrieved evidence. On questions derived from subsequent studies, both models also outscored expert-authored answers. In an experimental test of self-assembled monolayer design, OpenPerov selected a K analogue that outperformed the Na analogue preferred by experts. Devices incorporating the selected contact reached a power conversion efficiency of 27.05% and retained 96.74% of their initial efficiency after 1,000 h of dark thermal ageing at 85 $^\circ$C. These results show that reconstructed scientific reasoning provides an effective basis for building domain models that interpret experiments and inform materials design.
Large Language Models (LLMs) are increasingly used for code generation, raising concerns that they may be misused to produce malicious code. Meanwhile, Grammar-Constrained Decoding (GCD) has been widely adopted to improve the reliability of LLM-generated code by enforcing syntactic validity. In this paper, we reveal a counterintuitive risk: this reliability-oriented technique can itself become an attack surface. We uncover a new jailbreak attack, termed CodeSpear, that exploits GCD to induce LLMs into generating malicious code. Our experiments show that simply applying a benign code grammar constraint can effectively jailbreak LLMs. To address this vulnerability, we propose CodeShield, a safety alignment approach that robustly preserves safe behavior even under attacker-controlled grammar constraints. CodeShield aligns the model in the code modality by teaching it to generate honeypot code under GCD. Such code is semantically harmless, so it does not implement the malicious request, and structurally diverse, so it is difficult to suppress through grammar tightening. At the same time, CodeShield still preserves natural-language refusals when natural language is available. Experiments on 10 popular LLMs across 4 benchmarks show that CodeSpear outperforms representative jailbreak baselines and increases the attack success rate by more than 30 percentage points on average. CodeShield also restores safety under CodeSpear while preserving benign utility. Our findings reveal a fundamental risk of GCD and call for greater attention to its potential security implications.
We present an approach to mathematically formalizing PTX, an approach that allows modeling and reasoning about patterns common in GPU kernels for Deep Learning, including acquire-release, tensor cores, proxies, async operations, multi-kernel pipelines and data-dependent control flow. We implement it in Lean and use it to prove the correctness of simple implementations of a WGMMA GEMM kernel, FlashAttention, and a 3-kernel MoE pipeline.
Frontier language-model agents excel at complex tasks, but high inference costs make transferring their capabilities to smaller models essential. Traditional knowledge distillation achieves this through parameter updates, requiring access to the student's weights and additional training compute while encoding the transferred knowledge implicitly in the parameters. We introduce D-Skill, a framework that treats the teacher as a black box and distills its behavioral advantage into an interpretable natural-language skill by analyzing paired rollouts and identifying behavioral divergences and agreements. The resulting lessons are consolidated offline, with any compatible skill-update backend, into a skill document that is provided to the student at inference time. D-Skill constructs the skill entirely offline from fixed trajectories, without labels, further environment interaction, or weight updates after rollout collection. Across SpreadsheetBench-Verified, ALFWorld, and chess puzzles spanning multiple teacher-student pairs, D-Skill closes 75%-95% of the teacher-student gap without touching model weights. After a single offline pass over the paired traces, D-Skill outperforms one-epoch SFT baselines and remains competitive with multi-epoch SFT. Under the same model family, the resulting student also incurs 5.7x-8.3x lower inference cost per task than the teacher. Our results demonstrate that natural-language skill distillation offers a performant, interpretable, and efficient alternative to weight-based capability transfer.
Yuseon Choi, Jingu Lee, Jungjun Oh, Sunjoo Whang, Byeongcheol Kim, Minsung Kim, Junhyuk Lee, Hoi-Jun Yoo 等 9 位作者
Mixture-of-Experts (MoE) models have become the dominant architecture for large-scale language models, yet on-premises serving remains fundamentally memory-bound as batching turns sparse per-token compute into dense memory activation. Memory-centric architectures (PIM, NMP) improve bandwidth but leave compute underutilized under MoE's low arithmetic intensity on high batch sizes. Speculative decoding (SD) trades idle compute for fewer target invocations, yet verification must load experts even for rejected tokens, severely limiting its benefit in MoE especially on low batch sizes. We propose ELMoE-3D, a hybrid-bonding (HB)-based HW-SW co-designed framework that unifies cache-based acceleration and speculative decoding to offer overall speedup across batch sizes. We identify two intrinsic elasticity axes of MoE---expert and bit---and jointly scale them to construct Elastic Self-Speculative Decoding (Elastic-SD), which serves as both an expert cache and a strongly aligned self-draft model accelerated by high HB bandwidth. Our LSB-augmented bit-sliced architecture exploits inherent redundancy in bit-slice representations to natively support bit-nested execution. On our 3D-stacked hardware, ELMoE-3D achieves an average $6.6\times$ speedup and $4.4\times$ energy efficiency gain over naive MoE serving on xPU across batch sizes 1--16, and delivers $2.2\times$ speedup and $1.4\times$ energy efficiency gain over prior accelerator baselines.
In conversational memory, a later statement may replace an earlier fact, describe a different context, or be an isolated contradiction. Thus, agentic memory must track what changed over time without erasing what was true before. We introduce MINDSET, a memory controller that makes this distinction through explicit state transitions. It preserves original dialogue episodes and organizes them into versioned schemas. For each new episode, it chooses whether to reinforce a schema, supersede its current state, create a contextual branch, or start a new schema. A local energy score balances coherence, contradiction, historical preservation and fragmentation. Stability rules prevent isolated conflicts from immediately replacing established memory. On a frozen 700-question LoCoMo subset, MINDSET achieves the highest observed answer F1 among six evaluated systems and significantly improves retrieval ranking over LightMem. Without retuning, it obtains 69.23% judge accuracy on all LongMemEval knowledge-update cases, compared with 64.10% for Vector RAG. Two additional answer model backbones using Gemma-4-31B and GLM-4.7 provide further qualified evidence of generalization. These findings support explicit memory-state management as a useful approach to changing conversational knowledge.
Automated research pipelines pair an LLM proposer with a verifier on an implicit bet: discovery quality rises with verification strength. Under a pre-registered protocol we test this bet in two real-data domains -- job-shop scheduling and vehicular edge offloading on Beijing T-Drive trajectories -- across 100 matched LLM/random pairs per domain and three verifier tiers. The manipulation binds -- a 429-case mutation battery shows scores rising across tiers -- yet the paired LLM-over-random advantage does not scale with it (A, one-sided p=0.1115; D, p=0.9994), and in D shrinks under the strict tier (margin contrast -0.040, p=0.0013). The two lower gates promote identical candidates in 100/100 and 99/100 pairs, so the strict order is untestable when they coincide. Constructively, LLM proposals carry a real paired advantage over random controls, but a post-hoc union-pool analysis shows the fastest path to better discovery is widening the pool, not tightening the gate: mixing the two pools adds +1.52 holdout-value units in domain A (one-sided Wilcoxon, n=100, p=0.0002) -- evidence that pool composition, not verification strength, is the lever. Criteria, ledgers and scripts are archived on Zenodo (DOI: 10.5281/zenodo.23308866).
Memory enables large language models (LLMs) to use information beyond their parameters, such as past user conversations, long documents, or an agent's earlier actions. For a frozen LLM, access to this accumulated knowledge is limited to what a memory system places in its context window. However, gathering and combining relevant information needed for each query while keeping the model's context short remains challenging for existing systems. Here we introduce Memorilla, a memory system that gathers information from an entire document collection into a fixed number of query-dependent 'memory tokens' for a frozen LLM. Memorilla encodes documents (such as user chats, book chapters) once into reusable embeddings, then learns to retrieve and place query-specific information in the LLM's input embedding space. Using only 16 memory tokens, Memorilla achieves the best performance against baselines including Mem0 and MemoRAG on five of eight benchmarks spanning personalization, long-document and knowledge-intensive question answering. Notably, on FactKG, it reaches 84.33% accuracy versus 80.43% with full-context inference, while using 108x fewer tokens to represent the source documents. Furthermore, Memorilla can facilitate continual learning in agentic systems where the document count (the agent's past turns) increases throughout the trajectory. In a TextWorld environment, we show that Memorilla reduces the agent's tendency to wander between rooms it has already explored, and improves its average success rate from 18.6% with full-history in-context learning to 24.2%. Overall, Memorilla aids frozen LLMs in accumulating and using knowledge effectively across long-context interactions and takes a step towards building continually learning AI systems.
Web services increasingly expose backend interfaces that let users delegate tasks to AI agents, which invoke service operations directly without navigating websites. This shift changes how existing Web vulnerabilities can be exploited. We investigate these changes through VELA, an executable benchmark covering 44 vulnerabilities across twelve Web application families. The benchmark preserves vulnerable backends and derives agent-facing interfaces from legitimate workflows. We evaluate three commercial agents across multiple interface configurations, using controlled comparisons to separate interface effects from agent behavior and independent oracles to verify attack outcomes. Our results show that the agent realizes every attack that direct interface invocation realizes in the reference configuration, so exploitability is set mainly by the interface: 91% of vulnerabilities remain exploitable under the reference design. Interface restrictions prevent some attacks by removing required operations or limiting input control, while retention remains high across vulnerability classes. No case-level loss in the reference configuration is attributed to agent refusal or input normalization, and in one case the agent completed an attack by supplying calls the attacker omitted. Across the tested configurations, interface design affects exploitability more than agent choice, moving retention from 39% to 95% against at most 4.5 points across agents, but some restrictive interfaces also limit legitimate workflows. Our study provides a systematic empirical account of how agent-mediated access reshapes existing Web vulnerabilities, grounding future Web architecture design in measured security outcomes and trade-offs between reducing attack exposure and preserving legitimate functionality in agent-facing services.
Zhenbang Zhang, Zhengquan Luo, Kai Zhao, Zhiqiang Xu
2D Gaussian representations provide an explicit and lightweight alternative for image representation, enabling fast deterministic decoding without a heavy neural synthesis network. However, existing per-image Gaussian codecs still require costly image-specific optimization to determine how limited Gaussian capacity should be distributed across heterogeneous regions. We introduce CaliSplat, an optimization-efficient Gaussian image codec that selectively explores candidate representations and globally allocates a constrained Gaussian budget. We exploit the closed-form reconstruction error of the single-Gaussian case as an inexpensive measure of local difficulty, fitting higher-capacity candidates only in regions where additional Gaussians are likely to be useful. To reduce selection errors caused by mismatched convergence rates across candidate sizes, a small set of longer-horizon fits is used to calibrate short-horizon distortion estimates before global budget-aware selection. The selected representation is then compressed into a bitstream. Across standard and large image datasets, CaliSplat achieves competitive or superior rate--distortion performance while reducing per-image fitting time by $3.0\times$--$15.7\times$ compared with representative Gaussian codecs. It further maintains high-throughput deterministic decoding at $175.9$--$478.2$~MP/s. These results demonstrate that content-adaptive Gaussian compression can substantially reduce per-image optimization cost while preserving strong reconstruction quality and efficient explicit decoding.
Unsafe deserialization turns untrusted data into attacker-controlled execution, and a single vulnerability can admit diverse exploit variants that evade defenses tailored to a known exploit. Existing defenses typically operate at a single enforcement point with limited visibility into the full attack. Combining controls across layers offers complementary protection, but coordinating them to preserve legitimate functionality and adapting them to new variants requires substantial expert effort. We propose JANUS, a red-blue adversarial framework for automated cross-layer defense synthesis. Starting from a known exploit, \sys iteratively refines deployment-specific policies in an isolated offline environment. A black-box attacker probes the current policy for exploit variants using only externally observable feedback. The defender summarizes each verified variant into its security-relevant effects, uses counterfactual replay to separate complementary coverage from redundant enforcement, and jointly selects and scopes controls across layers. Each candidate policy is validated against accumulated exploits and benign workflows. We evaluate JANUS on sixteen vulnerability deployments across six Java and PHP applications against gateway, application, system, and cross-layer defenses. JANUS reduces held-out attack success to 6.2%, compared with 18.8-55.0% for the baselines, fully mitigates 13 of 16 deployments, and preserves 98.4% of benign functionality tests at a 6.9% targeted P95 latency increase.
Knowledge tracing is a key component of intelligent educational systems. However, most existing methods mainly focus on response prediction, providing limited support for explaining predicted outcomes and improving students' learning states. This paper proposes Diagnostic and Interpretable Knowledge Tracing (DIKT), which integrates response prediction, skill-level diagnosis, and model-based learning analysis. DIKT generates bounded cognitive-state trajectories and decomposes the final response probability into skill-wise probabilities. The trajectories reveal students' evolving strengths and weaknesses, while the skill-wise probabilities identify the skills that most strongly constrain the prediction. Based on these diagnostic results, DIKT supports direct cognitive-state intervention and effective-learning-duration-based what-if analysis to examine how hypothetical adjustments affect current predictions and subsequent state evolution. DIKT combines Bayesian multi-head attention, conditional variational recurrent inference, a logistic-growth decoder, and multiplicative skill-wise prediction. Experiments on three benchmark datasets show that DIKT outperforms all evaluated baselines, with an average absolute AUC gain of 2.04 percentage points. Other experiments further demonstrate that DIKT supports educationally meaningful diagnostic analysis and provides informative diagnostic indicators for downstream applications, extending knowledge tracing from response prediction to interpretable diagnosis and model-based learning analysis.
Zhijie Zhong, Zhiwen Yu, Kaixiang Yang, Yongheng Liu, Jun Jiang, C. L. Philip Chen
Time series anomaly detection (TSAD) research predominantly focuses on unsupervised architectures to model normal data dynamics under label scarcity. However, practical deployments frequently archive historical operational incidents, questioning whether continuous unsupervised architectural scaling remains the most cost-effective pursuit. In this paper, we empirically quantify the trade-offs between supervisory signals and model complexity. We introduce STAND, a deliberately minimalist supervised baseline based on bidirectional recurrent encoding. Across five benchmark datasets evaluated under a strictly disjoint out-of-sample temporal protocol, minimalist supervised models achieve competitive or superior accuracy relative to complex unsupervised alternatives. Furthermore, controlled label-scaling and few-shot evaluations confirm that these supervisory advantages persist under constrained historical event budgets, while STAND exhibits higher prediction certainty and sharper event delineation than reconstruction-based baselines. These findings encourage the TSAD community to complement algorithmic complexity with data-centric label utilization when annotations are accessible. The source code, benchmark datasets, and an extended Supplementary Document are publicly available at https://emorzz1g.github.io/STAND.