Update README.md

README.md

@@ -3,80 +3,194 @@ license: cc-by-nc-4.0
 task_categories:
   - video-text-to-text
   - visual-question-answering
-language: [en]
-tags: [video, temporal-grounding, one-to-many, benchmark, evaluation, multimodal, mllm]
+language:
+  - en
+tags:
+  - video
+  - temporal-grounding
+  - one-to-many
+  - benchmark
+  - evaluation
+  - multimodal
+  - mllm
 pretty_name: OMTG Bench
-size_categories: [n<1K]
+size_categories:
+  - n<1K
 ---
 
 # OMTG Bench: A Benchmark for One-to-Many Temporal Grounding
 
-OMTG Bench is the first comprehensive benchmark tailored for One-to-Many Temporal Grounding,
-introduced in **"Towards One-to-Many Temporal Grounding"** (ICML 2026, under review).
+**OMTG Bench** is the first comprehensive benchmark tailored for the **One-to-Many Temporal Grounding (OMTG)** task, introduced in the paper **"Towards One-to-Many Temporal Grounding"** (ICML 2026, under review).
+
+Unlike conventional temporal grounding benchmarks that assume a one-to-one mapping between a query and a single temporal segment, OMTG Bench evaluates a model's ability to retrieve **all** disjoint temporal segments in a video that correspond to a single textual query — a setting that is ubiquitous in real-world video content but on which state-of-the-art MLLMs (including Gemini-2.5-Pro, Gemini-3-Pro, Seed-1.8, and Qwen2.5-VL / Qwen3-VL families) exhibit a critical capability gap.
 
 ## Dataset Summary
 
 | Item | Value |
 |---|---|
+| Task | One-to-Many Temporal Grounding (OMTG) |
 | # Samples | 340 |
-| Annotation | Manually curated, expert-verified |
-| Consistency | > 90% |
-| Source video pools | Charades, ActivityNet, QVHighlights, VTimeLLM, Moment-10M (test only) |
-| Overlap with training | None |
-| Domains | Sports, cooking, news, etc. |
+| Annotation | Manually curated & verified by independent experts |
+| Inter-annotator consistency | > 90% |
+| Source video pools | Charades, ActivityNet, QVHighlights, VTimeLLM, Moment-10M (test splits only) |
+| Overlap with training set | None (by construction) |
+| Domains | Sports, cooking, news, and other diverse real-world scenarios |
+| Language | English |
+
+### Key Distribution Statistics
 
-### Distribution
 | Property | Value |
 |---|---|
-| GT segments per query | 2 – 20 |
-| 2–3 segments | 62.2% |
-| > 6 segments | 15% |
-| Video duration | 21 s – 17 min (avg 221.6 s) |
+| Ground-truth segments per query | 2 – 20 |
+| 2–3 segments per query | 62.2% of samples |
+| > 6 segments per query | 15% of samples |
+| Video duration | 21 s – 17 min |
+| Average video duration | 221.6 s |
+
+This distribution makes OMTG Bench a severe test of both **event cardinality perception** and **fine-grained temporal localization**, covering short clips as well as long-form narratives.
+
+## Supported Tasks
+
+- **One-to-Many Temporal Grounding** — given a video `V` and a textual query `Q`, predict the set `{[s₁, e₁], [s₂, e₂], …, [sₖ, eₖ]}` of all disjoint intervals where `Q` occurs.
 
 ## Evaluation Metrics
+
+The benchmark introduces a rigorous metric suite that jointly measures precision, recall, and event-count correctness:
+
 | Metric | Description |
 |---|---|
-| tIoU | Average temporal IoU over unions |
-| C-Acc | Count Accuracy (cardinality correctness) |
-| tF1@ξ | Temporal F1 at IoU ξ ∈ {0.3, 0.5, 0.7} via optimal bipartite matching |
-| **EtF1** | Effective Temporal F1 — strictly penalizes wrong cardinality (primary metric) |
+| **tIoU** | Average temporal IoU over unions of predicted and ground-truth intervals |
+| **C-Acc** | Count Accuracy — the percentage of samples whose predicted cardinality equals the ground-truth cardinality |
+| **tF1@ξ** | Temporal F1-Score at IoU threshold `ξ ∈ {0.3, 0.5, 0.7}` under optimal bipartite matching |
+| **EtF1** | **Effective Temporal F1** — assigns zero credit to any sample with incorrect predicted cardinality, strictly penalizing under- and over-retrieval and hallucinations |
+
+EtF1 is the recommended primary metric, as it couples instance-level precision–recall with event-count correctness.
+
+## Baseline Results on OMTG Bench
+
+A comprehensive assessment of representative open-source and proprietary MLLMs. All numbers are percentages (%).
 
-## Baseline Results (OMTG Bench)
 | Model | C-Acc | tF1@0.3 | tF1@0.5 | tF1@0.7 | tIoU | EtF1 |
 |---|---|---|---|---|---|---|
-| Qwen2.5-VL-3B | 0.00 | 15.17 | 7.01 | 2.86 | 11.60 | — |
-| Gemini-3-Pro | 30.63 | 58.30 | 47.75 | 29.89 | 47.63 | 21.30 |
-| Gemini-2.5-Pro | 50.94 | 55.72 | 43.57 | 27.97 | 43.24 | 27.80 |
+| **Proprietary Models** | | | | | | |
 | Seed-1.8 | 38.12 | 67.13 | 54.67 | 38.79 | 56.81 | 28.04 |
-| **OMTG-4B (ours)** | — | — | — | — | — | **43.65** |
+| Gemini-2.5-Pro | 50.94 | 55.72 | 43.57 | 27.97 | 43.24 | 27.80 |
+| Gemini-3-Pro | 30.63 | 58.30 | 47.75 | 29.89 | 47.63 | 21.30 |
+| **Open-Source General MLLMs** | | | | | | |
+| Qwen2.5-VL-3B | 0.00 | 15.17 | 7.01 | 2.86 | 11.60 | 0.00 |
+| Qwen2.5-VL-7B | 0.00 | 21.04 | 12.08 | 7.14 | 20.35 | 0.00 |
+| Qwen2.5-VL-32B | 0.00 | 16.81 | 9.66 | 4.76 | 18.32 | 0.00 |
+| Qwen2.5-VL-72B | 0.00 | 21.16 | 12.20 | 6.88 | 20.02 | 0.00 |
+| Qwen3-VL-4B | 0.31 | 37.07 | 26.75 | 17.93 | 30.42 | 0.21 |
+| Qwen3-VL-8B | 0.00 | 37.73 | 27.02 | 18.70 | 30.62 | 0.00 |
+| Qwen3-VL-30B | 0.00 | 37.03 | 25.98 | 17.52 | 32.36 | 0.00 |
+| Qwen3-VL-235B | 0.31 | 34.66 | 25.25 | 16.45 | 25.56 | 0.21 |
+| **Temporal-Grounding Experts** | | | | | | |
+| VideoChat-R1-7B | 0.00 | 32.07 | 19.70 | 10.42 | 24.93 | 0.00 |
+| VideoChat-R1.5-7B | 0.31 | 28.41 | 15.53 | 9.85 | 27.96 | 0.10 |
+| Time-R1-7B | 0.00 | 28.94 | 18.73 | 10.00 | 24.11 | 0.00 |
+| UniTime | 0.00 | 35.27 | 30.15 | 23.58 | 37.12 | 0.00 |
+| Timelens-8B | 0.00 | 39.14 | 32.76 | 22.58 | 32.38 | 0.00 |
+| **Ours** | | | | | | |
+| **OMTG-4B** | **55.63** | **73.46** | **65.40** | **48.96** | **61.24** | **43.65** |
+
+### Performance Gain of OMTG-4B across Training Stages
+
+| Stage | C-Acc | tF1@0.3 | tF1@0.5 | tF1@0.7 | tIoU | EtF1 |
+|---|---|---|---|---|---|---|
+| Base (Qwen3-VL-4B) | 0.31 | 37.07 | 26.75 | 17.93 | 30.42 | 0.21 |
+| + SFT (on OMTG-56K) | 44.06 | 69.57 | 61.23 | 45.63 | 56.94 | 34.81 |
+| + RL (GRPO) | **55.63** | **73.46** | **65.40** | **48.96** | **61.24** | **43.65** |
+
+Key observations:
+
+- Standard open-source models (Qwen2.5-VL series) yield **0% C-Acc / 0% EtF1**, failing to capture the one-to-many complexity.
+- Even the newer Qwen3-VL family (4B → 235B) barely exceeds 0% on cardinality-sensitive metrics.
+- Advanced proprietary models (Seed-1.8, Gemini-2.5/3-Pro) reach only **21–28% EtF1**.
+- Our **OMTG-4B** achieves a new state of the art at **43.65% EtF1**, outperforming the best proprietary baseline (Seed-1.8) by **+15.61 EtF1**.
 
 ## Dataset Structure
+
+### Files
+
 ```
 omtg_bench/
-├── OMTGBench.tsv   # 340 samples
-└── videos.zip      # ≈3.74 GB
+├── OMTGBench.tsv   # 340 annotated samples (74.2 kB)
+└── videos.zip      # corresponding source videos (≈3.74 GB)
 ```
 
-### Data Fields
+### Data Fields (`OMTGBench.tsv`)
+
+Each row is one OMTG evaluation sample. Typical columns:
+
 | Field | Type | Description |
 |---|---|---|
-| video_id | string | Video identifier |
-| duration | float | Video length (s) |
-| query | string | Natural-language query |
-| segments | list[[float,float]] | GT `[start,end]` intervals |
-| source | string | Charades / ActivityNet / QVHighlights / VTimeLLM / Moment-10M |
-| domain | string | Topical domain |
+| `video_id` | string | Unique video identifier (maps to a file in `videos.zip`) |
+| `duration` | float | Video length in seconds |
+| `query` | string | Natural-language query describing a repeating event |
+| `segments` | list[[float, float]] | Ground-truth `[start, end]` timestamps in seconds for all occurrences |
+| `source` | string | Originating dataset (Charades / ActivityNet / QVHighlights / VTimeLLM / Moment-10M) |
+| `domain` | string | Topical domain (e.g., sports, cooking, news) |
 
 ## Usage
+
 ```python
 from datasets import load_dataset
+
 bench = load_dataset("insomnia7/omtg_bench", split="train")
+print(bench[0])
 ```
 
+Videos can be extracted from `videos.zip` and matched via `video_id`.
+
+### Recommended Evaluation Protocol
+
+1. For each sample, prompt the model with the video and the textual query.
+2. Parse the model output into a set of `[start, end]` intervals (seconds).
+3. Compute **C-Acc**, **tF1@{0.3, 0.5, 0.7}**, **tIoU**, and **EtF1** via optimal bipartite matching against the ground-truth segments.
+4. Report **EtF1** as the primary metric.
+
+## Construction & Quality Control
+
+Samples were **randomly sampled and then manually curated** from the *test* splits of Charades, ActivityNet, QVHighlights, VTimeLLM, and Moment-10M. Care was taken to ensure:
+
+- **Zero overlap** between benchmark videos and any training data (including the companion [OMTG-56K](https://huggingface.co/datasets/insomnia7/omtg56k) dataset).
+- **Precise boundary annotations** verified by independent experts.
+- **Inter-annotator consistency > 90%**.
+
+## Intended Use
+
+- Evaluating MLLMs on one-to-many video temporal grounding.
+- Studying event cardinality perception and hallucination in video-language models.
+- Diagnosing over- and under-retrieval behaviors in long-form video understanding.
+
+## Limitations
+
+- Source videos inherit biases and licenses from the original datasets.
+- Domain coverage, while diverse, is not exhaustive; extremely long videos (> 20 min) are underrepresented.
+- Manual annotations, although expert-verified, may contain rare residual noise.
+- Intended for non-commercial research only.
+
 ## Companion Dataset
-Training dataset: [insomnia7/omtg56k](https://huggingface.co/datasets/insomnia7/omtg56k) (46k SFT + 10k RL).
+
+- Training dataset: [insomnia7/omtg56k](https://huggingface.co/datasets/insomnia7/omtg56k) — 56k high-quality instruction-tuning samples (46k SFT + 10k RL).
 
 ## License
-CC BY-NC 4.0 — non-commercial research only.
 
-```
+Released under **CC BY-NC 4.0** for non-commercial research use. Source videos remain subject to their original dataset licenses.
+
+## Citation
+
+```bibtex
+@inproceedings{omtg2026,
+  title     = {Towards One-to-Many Temporal Grounding},
+  author    = {Anonymous Authors},
+  booktitle = {Proceedings of the International Conference on Machine Learning (ICML)},
+  year      = {2026},
+  note      = {Under review}
+}
+```
+
+## Contact
+
+For questions or issues, please open a discussion on the dataset's Hugging Face page.