搜索结果：GPT-55

Authoring and refining presentation slides is a highly time-consuming core task in academic and business domains. While generative AI tools have lowered the barrier for creating initial drafts, their "black-box, one-way generation" paradigm severely deprives users of fine-grained control. Through a formative study (N=10), we identified "trial-and-error anxiety" and "inconsistent cross-page formatting" as primary bottlenecks in human-AI co-creation. Consequently, we present ECHO, an interactive system based on multimodal intent grounding and explainable operation plans. ECHO enables precise local edits via a "natural language + visual selection" paradigm, utilizing a decoupled "Plan-Confirm-Execute" loop and dynamic memory mechanisms to transform implicit AI intents into highly controllable layout co-creation. To systematically evaluate document refinement, we propose the CoEdit-Eval framework. Objective evaluations across multiple foundation models (e.g., GPT-5, GLM-4.7) demonstrate that while baselines uniformly fail in intent mapping (0% accuracy) and spatial grounding (0% Hit@1), the ECHO architecture boosts Target Hit@1 to 55%--85% depending on the base model. Furthermore, inte

Single-score translation metrics can conflate legitimate variation with error, a problem especially acute for classical languages where multiple defensible English renderings of the same passage coexist. We audit Pali-to-English output from four flagship large language models (LLMs): GPT-5.5, Claude Sonnet 4.6, Gemini 3.1 Pro, and Grok 4.3, on 1,700 passages from the Pali Canon, using three established human translations by Bhikkhu Sujato, Thanissaro Bhikkhu, and Bhikkhu Bodhi as a local reference envelope rather than a single gold standard. Each candidate's normalized embedding drift from the reference centroid serves as a triage signal, not an error label; the 1,203 candidates above a 1.5 drift threshold are then adjudicated by a blinded three-model LLM judge panel, calibrated against a 300-instance author-adjudicated validation set. Two results stand out. First, drift predicts severity rather than error per se: the major-error rate among adjudicated high-drift candidates rose monotonically from 7.9% in the 1.5-2.0 band to 51.6% above 3.0, while approximately 80% of 1.5-2.0 outliers were judged valid translation variations. Second, model differences were clearest in the high-drif

While skill optimization for autonomous agents has gained traction, existing methods rely on complex pipelines. This leaves a fundamental question unaddressed: What constitutes a minimal viable pipeline for skill optimization, where every component is justified by theory or empirical necessity? We formalize skill optimization via Zeroth-Order (ZO) optimization, mapping classical counterparts (central difference, trust regions) to recent literature. Noting that unlike blind numerical perturbations in classical ZO, skill trajectories serve as interpretable debugging feedback. Grounded in Claude Code philosophy and PAC learning, we establish three principles for convergence and generalization: file-system-based trajectory exploration, consensus attribute mining, and independent validation gating. Eliminating redundancies, we propose SkillOpt-Lite. It accelerates convergence and outperforms full SkillOpt: improving LiveMath by +8.8 points on GPT-5.5 and +25.4 points on GPT-5.4-nano, allowing the nano model to surpass standard GPT-5.4 optimized by SkillOpt. Finally, we integrate our framework into production coding agents like VSCode Copilot, enabling developers to evolve agent skills v

Real-world LLM applications are moving beyond single-agent workflows toward orchestrated multi-agent systems, yet current models still struggle to determine what each sub-agent needs to know. To measure this, we introduce PerspectiveGap, a benchmark for evaluating LLMs' ability to compose orchestration prompts for multi-agent systems. PerspectiveGap contains 110 scenarios, each evaluated through two distractor-mixed task formats: role-fragment assignment and free-form prompt writing. These scenarios are organized into 10 topologies, which are distilled from the authors' real-world engineering practice and framed by the Prompt Economy principle: building loop-centered orchestrations that maximize utility with minimal role and engineering overhead. In experiments with 27 commercial models from 10 companies, GPT-5.5 substantially outperforms all competitors, whereas Opus 4.7 shows a notable weakness in orchestration prompting despite its strong coding performance. Nevertheless, PerspectiveGap remains challenging: the evaluated models achieve an average combined pass rate of only 14.9\% (GPT-5.5 62.0\%) and an average overall leakage rate of 246.5\% (a per-scenario information leak-eve

Developers create modern software applications (Apps) on top of third-party libraries (Libs). When library vulnerabilities are reachable through application code, the applications can be vulnerable to software supply chain attacks. Prior work shows that developers often require concrete and executable evidence, i.e., proof-of-vulnerability (PoV) tests, to decide whether a reported dependency vulnerability poses a practical security risk to their application. However, manually crafting such tests is challenging, and existing tool support is insufficient to automate the procedure. To streamline test generation, we created PoVSmith -- a new approach that combines call path analysis, exemplar test, code context, and feedback into multiple prompts to guide a coding agent (i.e., Codex) and a large language model (i.e., GPT) for test generation, execution, and assessment. We evaluated PoVSmith on 33 $\langle$App, Lib$\rangle$ Java program pairs, where each App depends on a vulnerable Lib. PoVSmith revealed 158 unique application-level entry points (i.e., public methods) calling vulnerable library APIs; 152 (96\%) of them were correctly found, together with the call paths properly recogniz

We introduce EpiBench, a verifiable benchmark for short-horizon epigenomics analysis. EpiBench evaluates whether agents can make well-defined analysis decisions from realistic workflow states and return deterministically gradable answers. The benchmark includes 106 evaluations across CUT\&Tag/CUT\&RUN, ATAC-seq, ChIP-seq, and DNA methylation workflows. Across 5,088 valid trajectories from 16 model-harness pairs, no system passed a majority of attempts: GPT-5.5 / Pi led at 45.0\% (143/318 attempts; 95\% confidence interval (CI), 36.3--53.7), followed by GPT-5.5 / OpenAI Codex at 39.9\% (127/318 attempts; 95\% CI, 31.6--48.3). Claude Opus 4.8 Max / Pi and GPT-5.4 / Pi each passed 39.0\% (124/318 attempts; 95\% CI, 30.2--47.8 and 31.0--47.0, respectively). Performance varies across assay types, and many failed runs still contain parts of the correct answer. Agents often found the right files and computed useful intermediate results, but failed when the task required deeper, assay-specific scientific judgment.

As AI agents become increasingly capable of complex, long-horizon reasoning, rigorous and holistic evaluation is essential for measuring progress toward real-world healthcare applications. We introduce HealthAgentBench, a suite of 54 agentic healthcare tasks across 7 categories each with its unique environment. The benchmark suite spans diverse workflows throughout the patient journey and a broad range of modalities. Each task is designed to replicate an end-to-end clinical workflow: given minimal instructions, an agent must explore raw healthcare data, operate within a complex environment, and execute multi-step solutions that go beyond naive prompting. A final task success rate is reported to provide a single, interpretable metric for HealthAgentBench overall performance for each agent. Evaluating frontier agents on HealthAgentBench, we find that overall task success rate remains low, underscoring the difficulty of the suite. The strongest and the most cost effective agent, Codex GPT-5.5, achieves only approximately 42% success rate. Beyond aggregate performance, HealthAgentBench reveals nuanced strengths and weaknesses across task categories. Frontier agents show promise in auto

Skill documents provide procedural knowledge to large-language-model agents at inference time. This article studies whether the presentation granularity of controlled skill knowledge changes downstream task success. The experiment uses a pinned SkillsBench version, a 30-task domain-balanced subset validated by official oracle runs, two reasoning-enabled model configurations, six skill conditions, and five trials per task-condition-model cell. Skill availability is the clearest empirical signal. Relative to no skill, skill conditions increase task-mean pass rate by 26.7 to 36.0 percentage points for GPT-5.5 and by 18.0 to 26.0 percentage points for DeepSeek V4-Flash. The final data contain 1,800 rows, with 900 rows for each model. The task is the inference unit. Five trials are aggregated within each task-condition-model cell before paired contrasts are estimated over 30 tasks. The primary presentation contrasts are smaller and uncertain. Low-abstraction guidance differs from high-abstraction guidance by +0.7 percentage points for GPT-5.5 and -6.7 percentage points for DeepSeek V4-Flash, with both 95% bootstrap confidence intervals crossing zero. Adding one worked example to medium-

Objective: To describe the design and early clinical evaluation of The Daily Dose (TDD), an LLM-driven, automated clinical summarization and clinical-trial identification system integrated into routine radiation oncology practice. Design: Mixed-methods evaluation using a cross-sectional, anonymous clinician survey administered after 1 month of system deployment. Exposure: Daily automated delivery of physician-specific email summaries generated using RadOnc-GPT, including patient schedules, concise EHR-derived clinical-status summaries, and automated identification of potentially relevant clinical trials for new or consult visits. Main Outcomes and Measures: Primary outcomes included self-reported usability, satisfaction, perceived usefulness, perceived impact on workflow, time savings, and intention for continued use. Internal consistency reliability was assessed using Cronbach's $α$. Results: Among 55 respondents, 52 (94.5\%) worked in radiation oncology, and 38 (69.1\%) were attending physicians. Most participants (83.6\%) reported using TDD daily or several times per week. Mean (SD) scores were 3.89 (1.04) for usability and satisfaction, 3.43 (1.24) for perceived usefulness, and

Public viral genome resources such as the National Center for Biotechnology Information (NCBI) Virus database are central to outbreak response, evolutionary analysis, vaccine design, and genomic surveillance. Yet many high-value retrieval workflows remain optimized for interactive use rather than deterministic, reproducible programmatic interfaces. This creates a challenge for Large Language Model (LLM)-based scientific agents, where errors in metadata interpretation, filtering logic, or retrieval can propagate into incorrect datasets. To evaluate agentic viral data retrieval, we built VirBench, a manually curated benchmark of 120 queries spanning diverse pathogens, taxonomic levels, and metadata filters. When autonomous AI systems, including Biomni, Claude, GPT, and Edison Analysis, were tasked with these queries without a dedicated retrieval layer, performance varied widely: mean accuracy ranged from 16.9% for Claude Sonnet 4 to 91.3% for GPT-5.5, with newer frontier models showing progress but residual errors remaining consequential. To address this, we built gget virus, a deterministic query framework that formalizes NCBI Virus-style filtering as a reproducible programmatic sys

The misuse of Java security APIs is a serious security problem in software development. Research in 2024 has shown that this problem is widespread in LLM-generated code. However, it remains unclear whether this phenomenon persists in current models and how external security knowledge affects it. This paper presents a scoped replication and extension of Mousavi et al.'s study on the Java Cryptography Architecture (JCA) and Java Secure Socket Extension (JSSE) APIs. We focus on two complementary settings: GPT-5.5 as a frontier proprietary coding model, and Llama-3.3-70B-Instruct as a strong open-weight model relevant to self-hosted deployment. The results show that although newer LLMs perform better in using Java security APIs, the problem of Java security API misuse has not been eliminated. External security knowledge substantially improves the measured outcome, but its effect is model-dependent. For Llama-3.3-70B-Instruct, secure code examples are the most effective single knowledge type. For GPT-5.5, explicit misuse patterns eliminate all detected security API misuses among valid programs in our benchmark, although some outputs remain invalid due to compilation errors or target-API

Evaluation language is typically treated as a fixed English default in agentic code benchmarks, yet we show that changing the judge's language can invert backbone rankings. We localize the Agent-as-a-Judge prompt stack to five typologically diverse languages (English, Arabic, Turkish, Chinese, Hindi) and evaluate 55 DevAI development tasks across three developer-agent frameworks and six judge backbones, totaling 4950 judge runs. The central finding is that backbone and language interact: GPT-4o achieves the highest satisfaction in English (44.72\%), while Gemini leads in Arabic (51.72\%, $p<0.001$ vs.\ GPT-4o) and Hindi (53.22\%). No single backbone dominates across all languages, and inter-backbone agreement on individual requirement judgments is modest (Fleiss' $κ\leq 0.231$). A controlled ablation further shows that localizing judge-side instructions, not just benchmark content, can be decisive: Hindi satisfaction drops from 42.8\% to 23.2\% under partial localization. These results indicate that language should be treated as an explicit evaluation variable in agentic benchmarks. Full requirement-level judgments and runtime statistics are released for reproducibility.

Büyükaşik et al. [Publ. Math. Debrecen, 2024] introduced a family of generalizations of Euler's totient function $\varphi(n)$, by setting $\varphi_k(n) = \sum_{a} a^k$ for $a \in [1, n]$ such that $(a, n) = 1$, with $\varphi_0(n) = \varphi(n)$. Letting $\mathcal{D}_{s} = \{ k \geq s : \forall n \geq 1 \ \varphi_s(n) \mid \varphi_k(n) \}$, Büyükaşik et al. proved that $\mathcal{D}_{s}$ is finite for each $s \geq 0$, and conjectured that $\mathcal{D}_{1} = \{ 1, 3, 15 \}$ and provided computations to support this conjecture. We succeed in proving this conjecture, using an argument based on our extensive interactions with GPT-5.5 Pro.

Run the same LLM agent on the same task twice: do you get the same behavior? We find the answer is often no. In a study of 3,000 agent runs across three models (Llama 3.1 70B, GPT-4o, and Claude Sonnet 4.5) on HotpotQA, we observe that ReAct-style agents produce 2.0--4.2 distinct action sequences per 10 runs on average, even with identical inputs. More importantly, this variance predicts failure: tasks with consistent behavior ($\leq$2 unique paths) achieve 80--92% accuracy, while highly inconsistent tasks ($\geq$6 unique paths) achieve only 25--60%, a 32--55 percentage point gap depending on model. We trace variance to early decisions: 69% of divergence occurs at step 2, the first search query. Our results suggest that monitoring behavioral consistency during execution could enable early error detection and improve agent reliability.

Deploying autonomous edge robotics in dynamic military environments is constrained by both scarce domain-specific training data and the computational limits of edge hardware. This paper introduces a hierarchical, zero-shot framework that cascades lightweight object detection with compact Vision-Language Models (VLMs) from the Qwen and Gemma families (4B-12B parameters). Grounding DINO serves as a high-recall, text-promptable region proposer, and frames with high detection confidence are passed to edge-class VLMs for semantic verification. We evaluate this pipeline on 55 high-fidelity synthetic videos from Battlefield 6 across three tasks: false-positive filtering (up to 100% accuracy), damage assessment (up to 97.5%), and fine-grained vehicle classification (55-90%). We further extend the pipeline into an agentic Scout-Commander workflow, achieving 100% correct asset deployment and a 9.8/10 reasoning score (graded by GPT-4o) with sub-75-second latency. A novel "Controlled Input" methodology decouples perception from reasoning, revealing distinct failure phenotypes: Gemma3-12B excels at tactical logic but fails in visual perception, while Gemma3-4B exhibits reasoning collapse even w

LLM-agent training pipelines routinely discard failed trajectories even though GPT-4o achieves only 14-20% on WebArena and below 55% pass@1 on ToolBench; even specialised systems at 50-65% leave the majority of trajectories unused. We introduce AgentHER, which recovers this lost signal by adapting Hindsight Experience Replay (HER) to natural-language agent trajectories: a trajectory that fails goal A is often a correct demonstration for an achievable alternative goal B. AgentHER realises this through a four-stage pipeline (failure classification, outcome extraction, LLM-guided relabeling with confidence gating, and data packaging) that converts discarded failures into SFT, DPO, and ShareGPT training data. On WebArena and ToolBench under a strict task-disjoint held-out protocol, AgentHER improves over success-only SFT by +7.6-11.4% across four model families (GPT-4o, Qwen2.5-72B/7B, LLaMA-3.1-8B), achieves 2x sample efficiency, and beats the strongest experience-centric baseline (Agent Workflow Memory) by +3.0-6.2%. Two robustness mechanisms, failure-severity weighting and cross-model multi-judge verification (gpt-4o-mini paired with Qwen2.5-72B-Instruct), reduce label noise from 5.

Search-engine date filters are widely used to enforce pre-cutoff retrieval in retrospective evaluations of search-augmented forecasters. We show this approach is unreliable across two major search engines: auditing Google Search's before: filter and DuckDuckGo's date-range filter, we find that at least one retrieved page contains major post-cutoff leakage for 71% of questions on Google and 81% on DuckDuckGo, and the answer is directly revealed for 41% and 55%, respectively. Using gpt-oss-120b to forecast with these leaky documents, we demonstrate inflated prediction accuracy (Brier score 0.10 vs. 0.24 with leak-free documents). We characterize recurring leakage mechanisms, including updated articles, related-content modules, unreliable metadata, and absence-based signals, and argue that date-restricted search on these engines is insufficient for credible retrospective evaluation. We recommend stronger retrieval safeguards or evaluation on frozen, time-stamped web snapshots.

The Engineering Reasoning and Instruction (ERI) benchmark is a taxonomy-driven instruction dataset designed to train and evaluate engineering-capable large language models (LLMs) and agents. This dataset spans nine engineering fields (namely: civil, mechanical, electrical, chemical, environmental, aerospace, materials, fire, and industrial engineering) and 55 subdomains, and is crossed with seven intent types (i.e., definition, explanation, calculation, comparison, design/synthesis, troubleshooting, and code-related) and three difficulty tiers (undergraduate, graduate, and professional), yielding 57,750 records with field/subdomain/type/difficulty metadata and solution formatting. We examined ERI via seven LLMs and report a statistically significant three-tier performance structure, with frontier models (GPT-5, Claude Sonnet 4, DeepSeek V3.1) achieving mean scores above 4.30 on a five-point scale, while mid-tier and smaller models exhibited progressively higher failure rates and steeper performance degradation on graduate-level questions. To address circularity concerns inherent in LLM benchmarks, we developed a convergent validation protocol that leverages cross-provider independe

Can generative agents be trusted in multimodal environments? Despite advances in large language and vision-language models that enable agents to act autonomously and pursue goals in rich settings, their ability to reason about safety, coherence, and trust across modalities remains limited. We introduce a reproducible simulation framework for evaluating agents along three dimensions: (1) safety improvement over time, including iterative plan revisions in text-visual scenarios; (2) detection of unsafe activities across multiple categories of social situations; and (3) social dynamics, measured as interaction counts and acceptance ratios of social exchanges. Agents are equipped with layered memory, dynamic planning, multimodal perception, and are instrumented with SocialMetrics, a suite of behavioral and structural metrics that quantifies plan revisions, unsafe-to-safe conversions, and information diffusion across networks. Experiments show that while agents can detect direct multimodal contradictions, they often fail to align local revisions with global safety, reaching only a 55 percent success rate in correcting unsafe plans. Across eight simulation runs with three models - Claude,