In this paper, we introduce Predicate-State Video Reasoning Segmentation (PS-VRS), where a query defines a dynamic predicate P(o,t) over objects and time rather than a fixed object identity. Existing methods typically treat the target as permanently active once identified, leading to over-emission of masks during vacant intervals and failure to handle target switching. To address this, we construct PS-Bench, a large-scale benchmark with 2,077 videos and 16,250 queries annotated with explicit active/vacant predicate states, temporal intervals, and pixel masks. We further propose Predicate-State Grounding (PSG), a State-Bind-Emit pipeline that: (1) predicts [ACTIVE]/[VACANT] state tokens; (2) aligns them to video frames via dense cross-modal similarity; and (3) emits masks via SAM2 only for active states. Experiments show that PSG outperforms existing baselines by +9.32 𝒥 ℱ on PS-Bench and significantly reduces false-positive mask emissions. Our results demonstrate the effectiveness of explicit predicate-state modeling for video reasoning segmentation with dynamic visual witnesses.

Browser-use agents automate everyday web tasks through structured DOM interfaces or vision-language models operating on page screenshots. However, web pages can change between observation and action execution due to advertisements, asynchronous loading, and timed updates, causing agents to act on stale assumptions. We frame this temporal mismatch as a time-of-check to time-of-use (TOCTOU) vulnerability in browser-use agents, where page changes during planning can alter the target, value, or validity of a planned action before execution. We introduce DYNWEB, a benchmark spanning synthesized and real-world websites, and evaluate 10 open-source agents as well as Codex’s computer-use feature across different observation and action modalities. All 11 evaluated systems exhibit at least one TOCTOU violation, showing that the vulnerability is not limited to a particular agent design. To reduce this exposure, we design a lightweight pre-execution validation mechanism that reduces the exploitable window from approximately 10 s to 0.13 s. Across three agent frameworks, the mechanism reduces the number of vulnerable DYNWEB cases to 0 out of 14. Adaptive attacks further show that stealthy single-mutation attacks succeed in at most 2.2% of trials, while event-driven timing reveals the limits of pre-execution validation.

We present CyberChainBench, a benchmark for evaluating LLM-based agents on smart contract security across three complementary tasks: vulnerability detection, exploit generation, and patch synthesis. Built from 541 real-world exploit incidents from DeFiHackLabs spanning 9 EVM chains, the benchmark provides end-to-end on-chain evaluation where agents interact with historical blockchain state through isolated evaluation environments orchestrated by Harbor, using tools to read code, trace transactions, and validate exploits on mainnet forks. Each case is anchored to a specific block and includes structured ground truth covering vulnerability type, localization, and attacker profit. Exploits are graded by economic impact on historical forks; patches are validated by replaying historical attacks and legitimate transactions as fail-to-pass test oracles on a proxy-upgradeable subset. We define a five-type vulnerability taxonomy and evaluate multiple agent–model configurations. Results reveal a clear difficulty gradient: the best configuration scores 37.5% on detection, 43.7% on exploitation, but only 23.4% on patching, with the top agent realizing $57.4M in total exploit profit across 200 post-cutoff cases at a cost of 2.39 per case.

Evaluating multimodal large language models (MLLMs) on long-form scientific papers requires measuring not only answer correctness, but also whether answers are supported by traceable cross-modal evidence. Existing answer-only metrics and synthetic “Needle-In-A-Haystack” tests mainly assess answer matching or isolated retrieval, leaving evidence-grounded support under-examined. We introduce “Fish-in-the-Ocean” (FITO), an evidence-chain evaluation lens for native scientific documents, and instantiate it with SIN-Data, a scientific interleaved corpus that preserves text–figure coupling. Built on SIN-Data, SIN-Bench comprises four evidence-auditing tasks: evidence discovery, hypothesis verification, grounded QA, and evidence-anchored synthesis. We further propose “No Evidence, No Score”, which credits predictions only when linked to verifiable anchors and diagnoses evidence quality through matching, relevance, and logic. Experiments on 10 MLLMs show a clear correctness–grounding gap: Gemini-3-pro obtains the best average overall score (0.566), while GPT-5 reaches the highest SIN-QA answer accuracy (0.767) but lower evidence-aligned scores. Code and data are available at https://github.com/IIGROUP/sin-bench .

We present AnyCap, a plug-and-play framework that brings instruction alignment to omni-modal captioning. Captions offer a unified language interface for multimodal learning, and users increasingly expect instruction-driven control over their content and style. Current caption models lack explicit instruction supervision and are weak at instruction following, while directly tuning them can degrade general language ability. Achieving instruction alignment in an omni-modal setting is harder still, as each modality calls for separate models and custom designs. To address these challenges, AnyCap leverages a multimodal revision-based refinement paradigm that refines uncontrolled captions from existing models to instruction-aligned ones, without re-training base models. By processing multi-modality features in a unified framework, it enables one model to serve images, videos, and audio. To address the lack of instruction-based data, we construct AnyCapData, a large-scale, high-quality corpus spanning three modalities with 28 well-designed instruction types, each with multiple instruction samples. For evaluation, we address the limitations of current metrics for instruction-oriented captioning by designing AnyCapEval. Its key insight is to decouple evaluation into content and style for fine-grained assessment. Extensive experiments show that on AnyCapEval and diverse public benchmarks, AnyCap consistently improves both caption quality and instruction adherence for both open-source and API-based models. Notably, AnyCap-8B boosts GPT-4o’s content scores by 46% and style scores by 12%. Our code and models will be made publicly available.