# GridProbe: Posterior-Probing for Adaptive Test-Time Compute in Long-Video VLMs

Source: [GridProbe: Posterior-Probing for Adaptive Test-Time Compute in Long-Video VLMs](<https://www.sota2.com/research/paper/6a02b4addbc836b87d824f22>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-05-15T06:00:03.992Z

Authors: Mohamed Eltahir, Lama Ayash, Ali Habibullah, Tanveer Hussain, Naeemullah Khan

Published: 2026-05-11T15:57:46.000Z

## Abstract

Long-video understanding in VLMs is bottlenecked by a single monolithic forward pass over thousands of frames at quadratic attention cost. A common mitigation is to first select a small subset of informative frames before the forward pass; common for training-free selectors via auxiliary encoder-space similarities. Such signals are capped by contrastive pretraining, which usually fails on reasoning-heavy queries (negation, cross-frame counting, holistic summarization). We propose GridProbe, an efficient training-free posterior-probing inference paradigm that scores evidence in answer space using a frozen VLM's own reasoning and then selects question-relevant frames adaptively, resulting in sub-quadratic attention cost with little to no accuracy loss. We arrange frames on a $K{\\times}K$ grid and run lightweight row R and column C probes, where each probe reads its peak posterior as a query-conditioned confidence. The outer product of R and C yields an interpretable importance map whose skewness and kurtosis drive Shape-Adaptive Selection, a closed-form rule that reliably replaces the fixed frame budget $M$ with a per-question $M\_{\\mathrm{eff}}$. We show empirically that $M\_{\\mathrm{eff}}$ tracks intrinsic question difficulty without ever seeing the answer, a sign of test-time adaptive compute. On Video-MME-v2, GridProbe matches the monolithic baseline within $1.6$ pp Avg Acc at $3.36\\times$ TFLOPs reduction, while on LongVideoBench it Pareto-dominates the baseline ($+0.9$ pp at $0.35\\times$ compute). Because the selector and QA models can be decoupled, pairing a small 2B selector with a stronger 4B or 8B QA is strictly Pareto-dominant over the 2B monolithic baseline (up to $+4.0$ pp at $0.52\\times$ compute, on average), with no retraining. Finally, the interpretability of the importance maps opens future avenues for behavioral diagnostics, grounding, and frame-selection distillation.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Long-form Video Question Answering on LongVideoBench (test)](<https://www.sota2.com/research/sota/long-form-video-question-answering-on-longvideobench-test.md>): Long-form Video Question Answering; LongVideoBench (test); Long Video Accuracy: 55.9.

- [Video Question Answering on Video-MME no subtitles v2 (test)](<https://www.sota2.com/research/sota/video-question-answering-on-video-mme-no-subtitles-v2-test.md>): Video Question Answering; Video-MME no subtitles v2 (test); Non-Linear Accuracy: 12.86.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2605.10762>)

[PDF](<https://arxiv.org/pdf/2605.10762.pdf>)

[Implementation](<https://github.com/mohammad2012191/GridProbe>)