1
Will Scaling Improve Social Simulation with LLMs?
探讨LLM缩放能否提升社会模拟的保真度，揭示当前范式的局限与挑战
arXiv:2607.02464v1 Announce Type: new Abstract: Large Language Model (LLM) social simulations are a promising research method, but they are not yet fa…
2
Timesynth: A Temporal Fidelity Framework for Health Signal Digital Twins
论文提出Timesynth框架，专为健康信号数字孪生设计，解决时间保真度问题，推动精准医疗建模。
arXiv:2607.00431v1 Announce Type: new Abstract: Forecasting models for health-signal digital twins must preserve the oscillatory, frequency, phase, an…
3
Simulated Customers Never Walk Away: Decision Fidelity of LLM User Simulators Measured Against Real Purchase Outcomes
用真实购买结果衡量LLM用户模拟器的决策保真度，为对话AI评估提供新标尺
arXiv:2606.20708v1 Announce Type: new Abstract: LLM-as-user-simulation has become core infrastructure for conversational AI: agent benchmarks (tau-ben…
4
A Mechanistic Understanding of Pronoun Fidelity in LLMs
深入剖析LLM代词保真失效的机制，揭示推理、重复与偏见间的复杂关系。
arXiv:2606.16407v1 Announce Type: cross Abstract: Faithful and robust pronoun use is important for fair and coherent generations, yet large language m…
5
Moment-Video: Diagnosing Temporal Fidelity of Video MLLMs on Momentary Visual Events
最新研究为视频多模态大模型的时间感知能力设计诊断基准，聚焦瞬时视觉事件，暴露模型在极短事件上的理解缺陷。
arXiv:2606.02522v1 Announce Type: cross Abstract: Video multimodal large language models (MLLMs) have made rapid progress on general and long-form vid…
6
Probing Embodied LLMs: When Higher Observation Fidelity Hurts Problem Solving
提升观察精度反而降低问题解决能力——这项研究挑战了具身LLM的传统认知，揭示保真度与推理之间的意外权衡。
arXiv:2605.20072v1 Announce Type: new Abstract: Large Language Models are increasingly proposed as cognitive components for robotic systems, yet their…
7
Fidelity Probes for Specification--Code Alignment
论文提出规范性代码对齐的保真度探测方法，为新模型验证提供新思路。
arXiv:2605.17246v1 Announce Type: new Abstract: We introduce fidelity probes: natural-language questions generated from a reference artifact with code…
8
Dimension-Level Intent Fidelity Evaluation for Large Language Models: Evidence from Structured Prompt Ablation
提出维度级意图保真度评估框架，通过结构化提示消融实验揭示LLM的意图还原与形式复制差异。
arXiv:2605.14517v1 Announce Type: cross Abstract: Holistic evaluation scores capture overall output quality but do not distinguish whether a model rep…