---
title: "《低频陷阱：视频语言模型在简单事件记账中失效》"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2608.06361v1"
permalink: "https://feed.astralor.com/items/itm_14b74d42e31267a18c81a4f3"
published: "2026-08-06T17:57:06.000Z"
score: 78
channel: "ai"
tags: ["arXiv","Gemini 3.6 Flash","事件计数","时间推理","视频语言模型"]
---

# 《低频陷阱：视频语言模型在简单事件记账中失效》

## AI 点评

> 这篇论文最值得关注的地方在于它把“数错”这件事拆到了时间戳级别，而不是只看最终答案。它对做视频理解评估和模型训练的人特别重要，因为结果表明增加采样率可能只是在“刷分”。我建议大家读一下它的轨迹审计设计，或许能启发下一代多模态评测的玩法。

## 速览

本研究提出基于可执行事件轨迹的参数化评估框架，在弹球撞击、视觉闪烁、类别状态转移三类受控视频任务中，对视频语言模型的事件计数能力进行细粒度诊断。在2190个视频中，研究系统改变事件数量N与频率F，发现模型存在阶段性时间失效：Gemini 3.6 Flash在0.5和1.0 Hz下可稳定数出最多12个持续状态转移，但完全无法可靠计数瞬态闪烁事件。高频高计数场景下，最终计数正确率仅0.2%，事件恢复率仅18.1%。增加采样率虽能将弹球准确率从19.6%提升至29.3%，但序列级一致率仅3.7%，说明额外帧会虚增分数而非忠实恢复事件。该研究将视频评估从聚合准确率转向可定位的时序推理诊断。

## 关键点

- 提出轨迹锚定的参数化评估方法，用可执行事件轨迹实现时间戳级审计，而非仅看最终答案。
- 揭示阶段式时间失效：持续状态转移可数至12个事件，瞬态闪烁事件则完全无法可靠计数。
- 高频高计数场景下最终计数正确率仅0.2%，事件恢复率18.1%，增加采样率会虚增分数但序列一致性仅3.7%。
- 不同提示策略增益有限，真实视频评估同样呈现低事件计数区域的成功集中现象。

## 全文

Real-world video benchmarks provide broad coverage, but their fixed clips entangle event count, rate, duration, and visual complexity, making failure modes hard to isolate. While existing programmatic benchmarks offer better control, they score only the final answer rather than auditing reported events against executable ground truth. To bridge this gap, we introduce trace-grounded parametric profiling for event counting in three controlled video tasks: bouncing-ball wall contacts, visual blinks, and categorical state transitions. Across 2,190 videos, we vary event count N and frequency F while holding rendering fixed. Each video includes an executable event trace for capability-surface estimation and timestamp-level evaluation. Our results reveal a staged temporal failure. At an 80% reliability threshold, Gemini 3.6 Flash reliably counts persistent state transitions up to 12 events at 0.5 and 1.0 Hz, yet demonstrates no reliable positive-count region for transient blinking events. Thus, event representation dictates whether a model initially accesses evidence -- a limitation that compounds as count and frequency increase. In the high-count, high-frequency regime, only 0.2% of final counts are correct and the model recovers just 18.1% of true events. To test if visual access is the primary bottleneck, we increase sampling rate. Although this boosts Bounce Ball accuracy from 19.6% to 29.3%, the reported sequence agrees with ground truth only 3.7% of the time. Extra frames can therefore inflate final scores without producing faithful event recovery. Different prompting strategies yield similarly limited gains, and real-world video evaluations show the same concentration of success at low event counts. Ultimately, trace-grounded profiling shifts video evaluation from aggregate accuracy metrics to a detailed diagnostic of where temporal reasoning fails.

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_14b74d42e31267a18c81a4f3