TemporalCook: Benchmarking Temporal and Procedural Reasoning in Multimodal Large Language Models
Bibliographic record
Abstract
Multimodal large language models (MLLMs) have demonstrated impressive capabilities in integrating visual and textual information, with visual question answering (VQA) serving as a central task for evaluation. However, existing VQA datasets primarily target inference from static visual cues or factual content, leaving temporal and procedural reasoning underexplored. While video question answering allows for temporal understanding by providing access to full video sequences, in many real-world scenarios only a single image is available. To address this gap, we introduce TemporalCook, a new benchmark constructed from the YouCookII instructional video dataset that requires models to predict procedural and temporal outcomes based on static images in the cooking domain. To further augment temporal reasoning capabilities, we supplement the benchmark with external instructional videos. We also present two retrieval-augmented generation (RAG) baselines, leveraging either curated video knowledge source or open-domain retrieval from online video resources. We report a benchmark scoreboard for leading commercial and open-source multimodal models on TemporalCook, and evaluate the effectiveness of retrieval-augmented generation baselines. TemporalCook and these baselines provide a foundation for future research in temporal VQA and open new directions for developing retrieval-augmented solutions in temporally grounded multimodal tasks. The benchmark and evaluation code are publicly available at https://github.com/mrzarei5/TemporalCook.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.006 | 0.025 |
| Meta-epidemiology (narrow) | 0.003 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.002 | 0.002 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.002 | 0.004 |
| Open science | 0.004 | 0.003 |
| Research integrity | 0.003 | 0.004 |
| Insufficient payload (model declined to judge) | 0.011 | 0.005 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".