CRAFT: Critic-Refined Adaptive Key-Frame Targeting for Multimodal Video Question Answering

TL;DR

CRAFT is a novel multimodal video question answering pipeline that dynamically selects keyframes, verifies claims iteratively, and consolidates evidence across heterogeneous video sources, achieving state-of-the-art results.

Contribution

It introduces CRAFT, a query-conditioned, critic-refined pipeline combining dynamic keyframe selection, multilingual ASR, and iterative claim verification for improved multi-video QA.

Findings

CRAFT achieves the best overall average score of 0.739 on MAGMaR 2026.

It attains a reference recall of 0.810 and citation F1 of 0.635.

The approach generalizes well to non-overlapping event queries beyond MAGMaR.

Abstract

Grounded multi-video question answering over real-world news events requires systems to surface query-relevant evidence across heterogeneous video archives while attributing every claim to its supporting source. We introduce CRAFT (Critic-Refined Adaptive Key-Frame Targeting), a query-conditioned pipeline that combines dynamic keyframe selection, per-video ASR with multilingual fallback, and a hybrid critic loop to iteratively verify and repair claims before consolidation. The pipeline integrates UNLI temporal entailment, DeBERTa-v3 cross-claim screening, and a Llama-3.2-3B adjudicator, with a final citation-merging stage that emits each fact once with all supporting source identifiers. On MAGMaR 2026, CRAFT achieves the best overall average (0.739), reference recall (0.810), and citation F1 (0.635). We further evaluate on a MAGMaR-style conversion of WikiVideo with 52 non-overlapping event queries, where CRAFT also performs strongly (0.823 Avg), showing that its claim-centric evidence aggregation generalizes beyond MAGMaR. Ablations show that atomic claims, ASR, and the critic loop drive the main gains over the vanilla query-conditioned baseline. Code and implementation details are publicly available at https://github.com/bhosalems/CRAFT.