Tag
论文提出 ChronoFake 基准与 DelSA（文档选择自更新智能体）框架，通过 LLM 提取事件时间线并结合重排序器筛选证据，以检测通过重排真实事件时间顺序构造的时序错乱假新闻，超越现有方法。
The paper proposes T-Mem, a temporal-structure-based memory framework for LLM agents in long-horizon multi-party conversations, using Temporal Evidence Discovery and Temporal Memory Structuring to recover locally concentrated evidence and reconstruct event progressions as an evidence forest, outperforming existing memory frameworks on two benchmarks.
EventLens 提出一种以视觉为中心的强化学习框架，通过事件结构恢复（事件分割、进程判别、多事件关系推理）来训练多模态大模型，实现可验证奖励的 GRPO 训练与多教师在线策略蒸馏，在 Qwen3-VL 上提升时序理解且降低训练成本。
This paper systematically evaluates how LLMs generalize across four dimensions—domain shift, adversarial perturbations, compositionality, and length increase—in time and event expression extraction, finding that inductive prompting generalizes most consistently while gains from scale, architecture, and other prompting strategies are uneven and dimension-specific.
SeKondBrain AI Labs introduces the Concept Lifecycle Model and Concept-Grounded Attention, evaluating graph-injected attention, temporal versioning, and epistemic status on QA benchmarks. The results support explicit temporal validity and epistemic status while showing that several graph-attention mechanisms fail controlled comparisons with disabled-mechanism baselines.
论文提出 Temporal Activation Injection (TAI)，通过分析 VideoLLM 中帧序反转引发的层间表征差异（temporal divergence vector），在推理时沿实测衰减曲线重新注入时序信息，无需训练即可在三个 VideoLLM 和四个基准上一致提升时序推理能力。
BridgeMem proposes a new method for temporal knowledge graph forecasting by modeling dyadic transition residuals, demonstrating consistent improvements over baselines on standard benchmarks.
This paper presents Signal2Symbol, a neuro-symbolic framework for explainable anomaly detection in physiological time-series data like ECG and EEG, leveraging symbolic tokenization and temporal reasoning to produce interpretable explanations of anomalous patterns.
SCoP introduces a constraint-centric framework for temporal knowledge graph question answering that externalizes temporal decisions to filter evidence and improve answer accuracy on benchmark datasets like MultiTQ and TimelineCronQ-R.
This paper introduces a benchmark for measuring hindsight bias in clinical temporal reasoning of large language models, showing that exposure to future data leads to bias while temporal masking reduces it without lowering accuracy.
TempCloze is a video cloze benchmark for evaluating visual temporal reasoning in Video-LLMs, requiring models to identify missing video segments from distractors to reduce linguistic shortcuts in semantics, alignment, and progression.
This paper proposes a composable trust infrastructure for manufacturing knowledge graphs, focusing on cross-system provenance, temporal reasoning, and decision traceability.
This paper investigates whether language models consistently encode the current year through associative and declarative tasks, finding that the current year is not consistently encoded due to differing mechanisms between tasks.
The paper introduces temporal dependency graphs extracted from legal documents to compute filing deadlines, showing a pipeline approach outperforms direct language model answering in accuracy.
The paper constructs a benchmark to evaluate LLMs on temporal legal reasoning, revealing biases towards applying the most recently enacted laws and an inverse relationship between general reasoning ability and temporal performance.
CRAFT is a new LLM framework for iterative refinement of temporal reasoning over clinical narratives, introducing a verifier-based feedback mechanism and the MedTempo benchmark for vaccine adverse-event reports.
MobileMem introduces a benchmark and framework for on-device AI systems to learn from year-long mobile experiences, focusing on temporal reasoning, knowledge updating, and preference inference.
ChronoVision is a multimodal framework that improves temporal reasoning in vision-language models by aligning visual logic with latent imagery, using a Reconstructive Visual Head, ROI Attention Locating module, and reinforcement learning. It introduces the Vbvr-VQA dataset and achieves SOTA accuracy on temporal tracking benchmarks.
GROVE is a training-free framework that grows a temporally stratified memory from continuous video streams, supporting both reactive QA and proactive assistance. It achieves state-of-the-art results on benchmarks like MM-lifelong and EgoServe.
ReMem introduces a dual-level memory-augmented keyframe selection framework for training-free long video understanding, achieving state-of-the-art zero-shot performance on multiple benchmarks.