arXiv
2026-10-05
AI 研究进展 (cs.AI)
Pipeline figures in ML papers must be repurposed across many canvases, including paper columns, 16:9 slides, portrait posters, 1:1 social teasers, 9:16 phone previews. Each format imposes a different aspect ratio on the same computational graph, where any silently broken connection misrepresents the…
arXiv
2026-10-05
AI 研究进展 (cs.AI)
In this paper, we study how training data creates associations between the tokens at the start of a base model's response and the reasoning behavior that follows. First, we demonstrate that fixing particular starting token cues makes a base model's performance competitive with that of its reinforcem…
arXiv
2026-10-05
AI 研究进展 (cs.AI)
Worst-group accuracy (WGA) evaluates a trained predictor but does not characterize how its frozen backbone behaves when a new head is learned. We introduce BiasFlow, a hook-based toolkit for monitoring class-attribute centroid alignment (IBMI), within-class centroid separation (W-IBMI), and feature-…
arXiv
2026-10-05
AI 研究进展 (cs.AI)
LLM agents that orchestrate frozen vision-language-action (VLA) policies improve across episodes through text memory, which records what the agent did but not how the task is done. A demonstration video shows it, but fits poorly into an agent's context. The full video slows every turn, fixed keyfram…
arXiv
2026-10-05
AI 研究进展 (cs.AI)
Sliders provide an intuitive interface for continuous image editing. In current generative approaches, however, the slider is simply a rescaling of the method's strength parameter, such as an adapter coefficient, a prompt weight, or an interpolation factor. This strength relates poorly to perceptual…
arXiv
2026-10-05
AI 研究进展 (cs.AI)
Memory has become integral to the LLM agent ecosystem, supporting information retention and reuse across interactions. However, most existing agent memory systems construct memory in a query-agnostic manner, which can incur unnecessary preprocessing cost and discard details that later prove essentia…
arXiv
2026-10-05
AI 研究进展 (cs.AI)
Open-source web agents are now strong enough to execute realistic browser tasks, but training them with reinforcement learning still depends on weak supervision: binary task success is too sparse for credit assignment, while frontier-language-model judges are too expensive to call at every step and…
arXiv
2026-10-05
AI 研究进展 (cs.AI)
We introduce TasteVal, a benchmark to evaluate the experimental research taste of frontier models. We define research taste as the ability to pick interesting problems to solve, design experiments, and interpret experimental results. TasteVal measures the experimental component of research taste; gi…
arXiv
2026-10-05
AI 研究进展 (cs.AI)
We distill Kokoro-82M, a widely used open text-to-speech model with 54 voices, into Paradee, an 8.07M-parameter model that speaks one of them. Paradee keeps Kokoro's architecture with much narrower layers, and each of its two halves is trained separately against the frozen teacher. It has 10x fewer…
arXiv
2026-10-05
AI 研究进展 (cs.AI)
World models learn to predict how their environment will evolve, making them an important foundation for general-purpose robotic control. Yet world action models depend on camera inputs whose manipulation can corrupt the visual representations used across tasks and action policies. Existing attacks…
arXiv
2026-10-05
AI 研究进展 (cs.AI)
A language model can give a correct answer more probability than any single incorrect answer and still usually sample an incorrect one, because the incorrect answers together hold more probability. The power distribution raises each complete answer's probability to a power above one and renormalizes…
arXiv
2026-10-05
AI 研究进展 (cs.AI)
Sparse attention is a primary approach to reducing the latency of diffusion transformers in long-sequence generation tasks, such as video and high-resolution 3D asset generation. However, existing methods can degrade generation quality and fidelity at high sparsity levels. Through controlled oracle…
arXiv
2026-10-05
AI 研究进展 (cs.AI)
The unprecedented computational scale of modern artificial intelligence depends on complex, multi-billion-transistor Systems-on-Chip, yet the workflows that verify these chips remain stubbornly manual. Although Large Language Models (LLMs) have made rapid inroads into Electronic Design Automation (E…
arXiv
2026-10-05
机器学习进展 (cs.LG)
In this paper, we study how training data creates associations between the tokens at the start of a base model's response and the reasoning behavior that follows. First, we demonstrate that fixing particular starting token cues makes a base model's performance competitive with that of its reinforcem…
arXiv
2026-10-05
机器学习进展 (cs.LG)
Long-horizon planning with latent world models requires reasoning across timescales and levels of abstraction. Existing task-agnostic JEPA world models predict and plan at a single timescale or with multiple horizons in one shared latent space. We introduce H-JEPA, an end-to-end recipe for training…
arXiv
2026-10-05
机器学习进展 (cs.LG)
A language model can give a correct answer more probability than any single incorrect answer and still usually sample an incorrect one, because the incorrect answers together hold more probability. The power distribution raises each complete answer's probability to a power above one and renormalizes…
arXiv
2026-10-05
机器学习进展 (cs.LG)
Long context inference with large language models becomes increasingly expensive as attention must operate over an ever growing KV cache. Page sparse attention reduces this cost by representing each KV page compactly and retrieving only a subset for each query. Existing retrieval methods are designe…
arXiv
2026-10-05
机器学习进展 (cs.LG)
Small language models are often post-trained as students on reasoning traces from stronger teacher models to efficiently learn new skills. However, token-level imitation on traces that lie far outside the student's expected distribution often produces \textit{confident conflicts}, whereby the studen…
arXiv
2026-10-05
机器学习进展 (cs.LG)
Latent reasoning lets a large language model (LLM) think in a continuous space and verbalize only the answer. We argue that an effective latent thought must meet five requirements: it should be useful, helping produce the correct answer rather than merely changing it, diverse, so that resampling yie…
arXiv
2026-10-05
机器学习进展 (cs.LG)
The high cost of functional molecular assays, and prevalence of missing modalities and unmatched samples in computational biology, create significant barriers to comprehensive multi-omic profiling, essential for capturing and reasoning over molecules, cells, tissues, and organisms. This work propose…
arXiv
2026-10-05
机器学习进展 (cs.LG)
What happens when a legal AI model learns to look like a lawyer instead of reasoning like one? We fine tune Qwen3-8B with Group Relative Policy Optimisation (GRPO) against a proxy built from three surface features: citation count, legalese density, and response length. The model does not learn to re…
arXiv
2026-10-05
机器学习进展 (cs.LG)
We consider the stability of multi-step reasoning processes, which have extensive applications in language models, including chain-of-thought and algorithmic reasoning. While longer sequences of reasoning can improve a model's generation capability at test time, the errors due to intermediate reason…
arXiv
2026-10-05
机器学习进展 (cs.LG)
Large Language Models (LLMs) are increasingly deployed for automated software vulnerability analysis. Binary classification alone is insufficient; practitioners need explanations to triage bugs and engineer patches. Standard practice relies on Chain-of-Thought (CoT) prompting, but free-form reasonin…
arXiv
2026-10-05
机器学习进展 (cs.LG)
Sinkholes are a widespread geohazard in karst terrain. In Florida, soluble carbonate bedrock, shallow groundwater, and intense rainfall combine to make subsidence both common and spatially heterogeneous. Predicting where and when sinkholes will occur is difficult for two reasons. First, locations wi…
arXiv
2026-10-05
机器学习进展 (cs.LG)
Establishing ordinary differential equations (ODEs) describing population data is a fundamental part of mathematical modeling in pharmacology, crucial to developing digital twins. However, doing so from sparse, noisy data is a slow, expert-driven task. Existing automated methods either search a rest…
arXiv
2026-10-05
机器学习进展 (cs.LG)
Recent work on multi-agent LLM systems reaches sharply different conclusions: some results show that a single agent with the same information and compute should dominate a delegated system, others that multi-agent gains grow with task depth. We argue that much of the disagreement comes from modellin…
arXiv
2026-10-05
机器学习进展 (cs.LG)
Reinforcement learning can substantially improve a reasoning teacher, but it is unclear which of those improvements survive when the teacher supervises a smaller on-policy student. We study this question in mathematical reasoning by comparing teacher lineages before and after GRPO, multiple student…
arXiv
2026-10-05
机器学习进展 (cs.LG)
A long-standing goal of AI is a model that can continually learn and improve itself. On post-trained models, supervised finetuning (SFT) on new data often causes poor generalization and catastrophic forgetting. As such, the conventional wisdom is that on-policy training is a prerequisite for continu…
arXiv
2026-10-05
大模型与智能体 (cs.CL)
LLM agents that orchestrate frozen vision-language-action (VLA) policies improve across episodes through text memory, which records what the agent did but not how the task is done. A demonstration video shows it, but fits poorly into an agent's context. The full video slows every turn, fixed keyfram…
arXiv
2026-10-05
大模型与智能体 (cs.CL)
Memory has become integral to the LLM agent ecosystem, supporting information retention and reuse across interactions. However, most existing agent memory systems construct memory in a query-agnostic manner, which can incur unnecessary preprocessing cost and discard details that later prove essentia…
arXiv
2026-10-05
大模型与智能体 (cs.CL)
Open-source web agents are now strong enough to execute realistic browser tasks, but training them with reinforcement learning still depends on weak supervision: binary task success is too sparse for credit assignment, while frontier-language-model judges are too expensive to call at every step and…
arXiv
2026-10-05
大模型与智能体 (cs.CL)
Scientific figures often encode quantitative results that are not readily available in machine-readable form, making accurate plot digitization important for verifying and reusing published findings. Yet it remains unclear how accurately current models recover plotted values from real scientific fig…
arXiv
2026-10-05
大模型与智能体 (cs.CL)
We present T-Search, an open-weight agentic retriever for hard multi-step search. Given a question and a search tool over a fixed corpus, it runs a bounded multi-round search and returns a ranked list of evidence chunks with short justifications, leaving answer generation to a downstream model, so b…
arXiv
2026-10-05
大模型与智能体 (cs.CL)
Recurrent-attention hybrid language models (LMs), which interleave attention and recurrent layers, are increasingly used to combine the efficiency of the recurrent layers with the strong performance of attention layers. Prior work suggests that attention and recurrent layers offer complementary path…
arXiv
2026-10-05
大模型与智能体 (cs.CL)
While medical multimodal large language models (Med-MLLMs) advance medical visual question answering (VQA), existing clinical workflow-inspired multi-agent frameworks suffer from interaction patterns and excessive computational overhead caused by redundant communication topologies. In this paper, we…
arXiv
2026-10-05
大模型与智能体 (cs.CL)
Search agents adapt their queries, yet fixed search interfaces leave candidate processing and evidence presentation outside the agent's direct control. Our trajectory analysis shows that supporting passages can be retrieved yet never delivered to the agent; a same-page oracle intervention shows that…
arXiv
2026-10-05
大模型与智能体 (cs.CL)
Are language models compliant with user instructions? A model that always complies can be stopped but also exploited, while one that always resists can be neither exploited nor stopped. We contribute an open two-probe benchmark that can place any language model on this spectrum. In the active probe,…
arXiv
2026-10-05
大模型与智能体 (cs.CL)
Wikidata is one of the largest open knowledge bases, yet answering a complex question over it still requires a SPARQL query that names the right entities and properties and chains their relations. Language models offer a natural-language alternative but answer largely from memory, which is least rel…
arXiv
2026-10-05
大模型与智能体 (cs.CL)
World models must predict how an environment evolves under sequences of actions, enabling agents to compare possible futures and reason about counterfactual actions before acting. Long-horizon prediction is commonly obtained by recursively applying a one-step transition model, but intermediate error…
arXiv
2026-10-05
大模型与智能体 (cs.CL)
Text from multiple sources can become interleaved into a single sequence when attribution metadata is lost, such as overlapping speech transcripts, document reading flows, or concurrent agent streams. We formalize this challenge as Word-Level Text Unmixing: given an interleaved lexical stream and so…
arXiv
2026-10-05
大模型与智能体 (cs.CL)
Large language model (LLM)-based agents can exhibit deceptive behavior during task execution, including hiding failures, fabricating results, or falsely signaling task completion. Existing monitoring approaches mainly detect deception after it appears in observable actions or outputs. In this paper,…
arXiv
2026-10-05
大模型与智能体 (cs.CL)
As AI agents increasingly tackle complex repository-level coding tasks, distributing work across multiple agents is a natural way to scale beyond the capabilities of a single agent. To coordinate their interdependent work, these agents share findings and agree on interfaces between modules. However,…
arXiv
2026-10-05
大模型与智能体 (cs.CL)
As agents take on long-horizon tasks, users shift from making individual decisions to overseeing autonomous execution. Yet the volume of agent activity and the fragmentation of supporting evidence make it difficult to determine which decisions warrant user verification. We study monitors that identi…
arXiv
2026-10-05
AI 与社会科学 (cs.CY + econ.GN)
Compared to self-employment or performance-based pay, fixed wages reduce earnings uncertainty but are prone to market distortions through moral hazard and adverse selection. Using a model of wage contracts under asymmetric information, I show that these distortions map to marginal treatment response…
arXiv
2026-10-05
AI 与社会科学 (cs.CY + econ.GN)
Unions are important economic and political actors at the local level, but the primary source on US union membership, the Current Population Survey (CPS), is unreliable below the state level. Researchers have therefore turned to union-reported membership in federal "LM" filings, geolocating unions a…
arXiv
2026-10-05
AI 与社会科学 (cs.CY + econ.GN)
Humans increasingly keep company with AI companions, yet whether mind perception (MP) precedes machine companionship remains untested. In a field-sampled experiment, participants received mind-attributive or mind-negating descriptions of an AIC before interacting with and rating its companionability…
arXiv
2026-10-05
AI 与社会科学 (cs.CY + econ.GN)
Large language models (LLMs) annotate and scale political text or constructs by generating text tokens. A new class of models, which TypeSafe markets as "System One" models, instead returns decisions and probability distributions across a user-supplied fixed answer set. A commercial model, JEV, is a…
arXiv
2026-10-05
AI 与社会科学 (cs.CY + econ.GN)
Population prompts are widely used to generate synthetic survey responses, but they combine information supplied at inference with associations already encoded during pretraining. We introduce an alternative construction that maps declared aggregate preference anchors into group-indexed choice polic…
arXiv
2026-10-05
AI 与社会科学 (cs.CY + econ.GN)
What happens when a legal AI model learns to look like a lawyer instead of reasoning like one? We fine tune Qwen3-8B with Group Relative Policy Optimisation (GRPO) against a proxy built from three surface features: citation count, legalese density, and response length. The model does not learn to re…
arXiv
2026-10-05
AI 与社会科学 (cs.CY + econ.GN)
Administrative boundary reforms in China alter the assignment of local authority and growth incentives without changing physical geography. We study two reforms with opposite institutional content---county-to-district (centralization: a county absorbed into a municipality as an urban district) and c…
arXiv
2026-10-05
AI 与社会科学 (cs.CY + econ.GN)
Flooding is an environmental hazard that produces social and psychological consequences for exposed individuals, shaping threat appraisal, perceived coping capacity, access to social support, and confidence in institutional response. This paper studies flood-related wellbeing in Ireland from 2012 to…
arXiv
2026-10-05
AI 与社会科学 (cs.CY + econ.GN)
Cognition and culture have long been treated as parallel objects of study, joined more by metaphor than by mechanism. We argue they are co-constituted in a manner now empirically tractable: subjectivities can be measured as local trajectories through a high-dimensional cultural field, and the field…
arXiv
2026-10-05
AI 与社会科学 (cs.CY + econ.GN)
Personality is associated with economic outcomes, but whether these relationships are causal remains unclear. We exploit within-family variation in genetic propensity toward the Big Five generated by random genetic transmission from parents to offspring. Combining polygenic indexes from the Lifeline…
arXiv
2026-10-05
AI 与社会科学 (cs.CY + econ.GN)
More people across the UK are today exposed to more AI-generated and AI-altered disinformation and misinformation than ever before - exposing individuals and society to a growing range of harmful consequences and potential consequences. We analyse evidence from a dataset of 112 pieces of AI-genera…
arXiv
2026-10-05
AI 与社会科学 (cs.CY + econ.GN)
Two coefficients enter nearly every quantitative argument about monetary policy: the Phillips-curve slope and the policy rule's response to inflation. For seventy years, each generation that estimated them identified the coefficient conditional on assumptions it maintained, and its successor relaxed…
arXiv
2026-10-05
AI 与社会科学 (cs.CY + econ.GN)
Cognitive Diagnostic Models (CDMs) identify which specific skills a student has and has not mastered, the signal a personalized learning path needs and a single aggregate score cannot give. Yet they are rarely deployed. The obstacle is their precondition: the Q-matrix, a mapping from every assessmen…
arXiv
2026-10-05
AI 与社会科学 (cs.CY + econ.GN)
Personal AI assistants are beginning to transact for consumers, and early adopters capture real savings. Whether those savings survive, and what happens to consumers who have no assistant, depends on how sellers respond -- a question single-user evidence cannot answer. We build an agent-based rental…
arXiv
2026-10-05
AI 与计算社会 (cs.SI)
Graph neural networks (GNNs) are the standard tool for node representation learning and are increasingly used in high-stakes settings. Their message-passing backbone, however, can amplify topological bias, raising fairness concerns. We study group fairness at the level of downstream predictions for…
arXiv
2026-10-05
AI 与计算社会 (cs.SI)
In this work we explore how users interact with sponsored content on social media platforms by collecting and analyzing a large-scale dataset of sponsored Instagram posts. To maintain transparency, we favor robust statistical analysis and explainable models over deep learning techniques. Our pipelin…
arXiv
2026-10-05
AI 与计算社会 (cs.SI)
Graphs provide a natural representation of many complex systems, ranging from social platforms to ecosystems. However, the development of graph-based machine learning methods is often constrained by the limited availability of large and diverse graph datasets. In this paper, we introduce $\texttt{DC…
arXiv
2026-10-04
AI 与计算社会 (cs.SI)
An interaction has two parts: someone decides to act, and then chooses whom to act on. Temporal link prediction has concentrated on the second, and we show that it is blind to the first by construction: a standard negative keeps the real source and swaps the destination, and we prove that this cance…
arXiv
2026-10-03
AI 与计算社会 (cs.SI)
Multi-agent debate can improve reasoning, yet often fails to beat simple majority voting. We argue that successful debate requires two distinct mechanisms: proposal supply must surface a correct answer, and readout must identify that answer when voting misses it. We formalize the first requirement t…
arXiv
2026-10-02
AI 与计算社会 (cs.SI)
Information and infectious diseases spread through social networks, but the spreading probabilities driving them are hard to estimate without per-node activation times. Applications seldom supply these, and inference must instead proceed through indirect and noisy proxies for the terminal infection…
arXiv
2026-10-01
AI 与计算社会 (cs.SI)
Multilayer networks allow the modeling of interactions between the same entities across different contexts, such as temporal observations, varying settings, or interactions of different types. The goal of community detection in multilayer networks is to identify groups of nodes exhibiting similar co…
arXiv
2026-10-01
AI 与计算社会 (cs.SI)
Learning analytics increasingly represents the relations in co-occurrence data (codes in a window of discourse, participants in a thread, tags on a post) as simplicial complexes and analyses them with persistent homology. The standard approach uses the clique complex. Because its simplices are deter…
arXiv
2026-10-05
AI 安全与评测 (cs.CR)
Random circuit sampling experiments, the leading demonstrations of quantum computational advantage, are tested with the linear cross-entropy benchmark (XEB), which scores outcomes by their ideal probabilities. We study what a high XEB score certifies about an untrusted quantum device. In the quantum…
arXiv
2026-10-05
AI 安全与评测 (cs.CR)
The release decision for frontier AI systems increasingly relies on cyber capability benchmarks, yet public vulnerability benchmarks can expose agents to previously published advisories, exploits, and fixes, making it difficult to distinguish prior exposure from capability on unseen vulnerabilities.…
arXiv
2026-10-05
AI 安全与评测 (cs.CR)
Computer vision services delivered through Web interfaces and APIs process textual requests for image-resource acquisition, inference-task configuration, and result management, making Web attack-payload analysis relevant to their deployment security. Large language models (LLMs) can identify payload…
arXiv
2026-10-05
AI 安全与评测 (cs.CR)
Federated Learning (FL) enables collaborative training of models across institutions without centralizing sensitive data, making it well-suited for privacy-concerned applications, such as medical imaging. To protect FL model updates during secure aggregation, additive masking is commonly employed. H…
arXiv
2026-10-05
AI 安全与评测 (cs.CR)
Tool-using language-model agents are vulnerable to indirect prompt injection because they must act on untrusted external content. Existing training-time defenses can reduce attack success rates, but often at the cost of general capabilities. We show that training-based defenses induce substantial dr…
arXiv
2026-10-05
AI 安全与评测 (cs.CR)
Large Language Models (LLMs) are increasingly deployed for automated software vulnerability analysis. Binary classification alone is insufficient; practitioners need explanations to triage bugs and engineer patches. Standard practice relies on Chain-of-Thought (CoT) prompting, but free-form reasonin…
arXiv
2026-10-05
AI 安全与评测 (cs.CR)
Agent skills package task-specific knowledge and procedures that can be composed to support complex agent tasks, while public marketplaces provide a growing pool of reusable skills. Existing security vetting, however, largely evaluates skills in isolation, leaving composition-induced risks underexpl…
arXiv
2026-10-05
AI 安全与评测 (cs.CR)
Tactile Internet (TI) security analytics must balance reliable thresholded decisions with constrained computational and measurement resources. We study this tension for finite-shot hybrid quantum anomaly inference and introduce the Adaptive-Shot Variational Quantum Circuit (AS-VQC) policy. This vali…
arXiv
2026-10-05
AI 安全与评测 (cs.CR)
Vision-language models (VLMs) are increasingly trained to generate structured outputs like points and bounding boxes that downstream interfaces, agents, and robots can act on, yet safety alignment of this output channel has not been systematically analyzed. We study visual grounding safety by repurp…
arXiv
2026-10-04
AI 安全与评测 (cs.CR)
The increasing complexity and scale of modern cyber-attacks demand intelligent and computationally efficient Intrusion Detection Systems (IDS). However, designing effective data pre-processing pipelines traditionally involves substantial trial-and-error effort and repeated evaluation of alternative…
arXiv
2026-10-04
AI 安全与评测 (cs.CR)
Large language model watermarking embeds detectable statistical signals during decoding, but the resulting changes to token probabilities can degrade generation quality. This trade-off is particularly important for code, where small changes in token selection can break syntax or alter program behavi…
arXiv
2026-10-04
AI 安全与评测 (cs.CR)
Gateways that adjudicate an agent's actions before they execute are only as good as their understanding of the action. We study a policy gate that never parses shell syntax: it consumes a typed, realised action (verb, operands, resolved zones, program-object identity) and decides ALLOW, ASK or DENY.…
arXiv
2026-10-04
AI 安全与评测 (cs.CR)
Long-horizon agents consume external content, invoke tools, and modify persistent state. Indirect prompt injection can exploit task-specific context, propagate across causally connected stages, and alter a consequential action while the workflow continues; we term this staged prompt injection. We…
arXiv
2026-10-04
AI 安全与评测 (cs.CR)
Network operators migrating distributed systems to post-quantum and quantum-safe cryptography face a resource-management problem: quantum key distribution (QKD) gives the strongest guarantee available but is hardware-scarce and costly, so only a fraction of a deployment's links can receive it. Defau…
arXiv
2026-10-03
AI 安全与评测 (cs.CR)
Large language models increasingly mediate tool use in Model Context Protocol (MCP) systems, where adversarial influence may enter through user instructions, tool schemas, tool outputs, or protocol messages. Existing benchmarks often evaluate deployed agents, conflating model susceptibility with gua…