Publish StreamOPD annotation release v1

Three privacy-sanitized annotation configurations, video-source documentation and local preparation tools. No videos or private logs.

LICENSE.md

@@ -0,0 +1,20 @@
+# Component terms and source attribution
+
+This release combines derived annotations and generated cue text with references
+to independently distributed videos. A single blanket license for all components
+is not asserted, so the dataset card uses `license: other`.
+
+- **ThinkStream QA source**: `CASIA-IVA-Lab/ThinkStream` declares MIT in its dataset
+  card. Its upstream notice is reproduced in `licenses/ThinkStream-NOTICE.md`.
+- **LLaVA-Video-178K video references**: no videos are redistributed here. The
+  inspected dataset card does not declare a single license in its metadata;
+  consult its source-component terms and original video rights.
+- **Tarsier2-Recap-585K / Kinetics-700**: the upstream dataset metadata declares
+  Apache-2.0 and the repository is access-gated. Obtain access and review the
+  upstream conditions directly. This release contains path references, not videos.
+- **Additional cue annotations, packaging and helper scripts**: no additional
+  standalone license grant is asserted by this initial release. Contact the
+  repository maintainer for reuse permissions not covered by the upstream terms.
+
+Do not apply the QA annotation license to third-party video content. Attribution
+and upstream access obligations continue to apply to the respective components.

README.md

@@ -0,0 +1,182 @@
+---
+pretty_name: StreamOPD Data
+license: other
+license_name: upstream-and-component-terms
+license_link: LICENSE.md
+language:
+- en
+task_categories:
+- video-text-to-text
+- question-answering
+tags:
+- streaming-video
+- video-question-answering
+- on-policy-distillation
+- visual-cues
+- annotations-only
+size_categories:
+- 10K<n<100K
+configs:
+- config_name: current
+  default: true
+  data_files:
+  - split: train
+    path: data/current/train.parquet
+  - split: validation
+    path: data/current/validation.parquet
+- config_name: legacy_25118
+  data_files:
+  - split: train
+    path: data/legacy_25118/train.parquet
+  - split: validation
+    path: data/legacy_25118/validation.parquet
+- config_name: grpo_25114
+  data_files:
+  - split: train
+    path: data/grpo_25114/train.parquet
+  - split: validation
+    path: data/grpo_25114/validation.parquet
+---
+
+# StreamOPD: Training Data for Streaming Video Post-Training
+
+StreamOPD-Data provides video question-answer annotations, local video time intervals,
+and optional teacher-side visual cues for on-policy distillation and related controls.
+**Video files, model weights, training logs and credentials are not included.**
+The QA annotations are derived from [ThinkStream](https://huggingface.co/datasets/CASIA-IVA-Lab/ThinkStream).
+Video references resolve to [LLaVA-Video-178K](https://huggingface.co/datasets/lmms-lab/LLaVA-Video-178K)
+and the Kinetics-700 component of [Tarsier2-Recap-585K](https://huggingface.co/datasets/omni-research/Tarsier2-Recap-585K).
+
+## Configurations
+
+| Configuration | Train | Validation | Training rows with cues | Purpose |
+|---|---:|---:|---:|---|
+| `current` (default) | 25,106 | 326 | 24,245 | Current shared OPD/CueGate pool after known input and source-overlap exclusions |
+| `legacy_25118` | 25,118 | 326 | 24,257 | Earlier QA pool with the explicit cue-instruction teacher prompt |
+| `grpo_25114` | 25,114 | 326 | 0 | Readability-filtered GRPO data with the actual direct-answer prompt suffixes |
+
+These are overlapping versions of the same data family, **not independent datasets to concatenate**.
+Validation questions, answers and intervals are shared; the GRPO version has its own
+direct-answer prompt wording. Each configuration's original row order is retained.
+The `legacy_25118` configuration retains the known problematic rows for historical
+reproduction; use `current` unless reproducing that specific configuration.
+
+Two source questions contained email text among answer choices. This publication replaces
+those strings with `[REDACTED_EMAIL]` in every affected prompt, without changing answer
+letters. `privacy_redacted` marks the two affected training rows in each configuration.
+Therefore this release is a privacy-sanitized representation, not a byte-identical copy
+of private training inputs. See [data processing](docs/data_processing.md).
+
+## Load annotations
+
+```python
+from datasets import load_dataset
+
+data = load_dataset("UniX-Lab/StreamOPD-Data", "current")
+print(data["train"][0]["prompt"])
+print(data["train"][0]["videos"])
+```
+
+`videos` contains plain relative-path metadata, not embedded video bytes or automatic
+downloads. Obtain the upstream videos using [video_sources.md](docs/video_sources.md).
+
+Download this small annotation release with one worker:
+
+```python
+from huggingface_hub import snapshot_download
+
+snapshot_download(
+    repo_id="UniX-Lab/StreamOPD-Data",
+    repo_type="dataset",
+    local_dir="StreamOPD-Data",
+    max_workers=1,
+)
+```
+
+Resolve paths into local trainer inputs after obtaining the videos:
+
+```bash
+python StreamOPD-Data/scripts/prepare_local_data.py \
+  --config current --video-root /path/to/videos \
+  --output-dir ./streamopd-local
+
+python StreamOPD-Data/scripts/validate_data.py
+```
+
+The preparation tool requires all video paths to resolve, preserves timestamps and
+annotation content, and writes `train.parquet` and `validation.parquet` to a new local
+directory. It supports the extra `data/` alias used in some historical LLaVA paths.
+It checks file presence without decoding videos; the known legacy interval failures
+remain documented rather than silently filtered.
+
+## Fields
+
+| Field | Meaning |
+|---|---|
+| `sample_id` | Stable ID, including split; shared across configurations for the same underlying source row |
+| `source_row_id` | Original zero-based row number in the base train pool or validation split |
+| `privacy_redacted` | Whether contact text was replaced in this row for publication |
+| `prompt` | Student chat messages, with `<video>` placeholder and question/choices |
+| `videos` | List of `{video, video_start, video_end}`; paths relative to the video root, times in seconds |
+| `video_source` | `LLaVA-Video-178K` or `Kinetics-700` |
+| `ground_truth` | Verifiable reference answer as a string |
+| `response_format` | `Multiple Choice`, `Binary`, or `Counting` |
+| `data_source` | Preserved training reward-routing tag; not the video's provenance |
+| `reward_model` | Reference-answer struct expected by the original training interface |
+| `teacher_prompt` | Cue-augmented teacher messages where available; otherwise identical to the student prompt |
+| `teacher_reference_raw` | Original student messages used as a no-cue reference |
+| `teacher_reference_empty` | Same cue instruction wrapper with the cue removed; or original messages on fallback rows |
+| `cue_text`, `cue_available` | Optional generated pointer and availability flag |
+
+Teacher-reference fields are stored for convenience in every configuration. Only
+`current` used both stored reference forms in the reference-template preparation;
+the corresponding fields in `legacy_25118` are reconstructed from the unchanged cue
+wrapper. All teacher fields in `grpo_25114` equal its direct-answer student prompt.
+Plain OPD must use the student prompt without injecting `teacher_prompt` as a cue.
+
+## Current training split statistics
+
+| Answer format | Rows |
+|---|---:|
+| Multiple choice | 17,485 |
+| Binary | 6,364 |
+| Counting | 1,257 |
+
+19,814 training records reference LLaVA-Video-178K and 5,292 reference Kinetics-700.
+There are 22,848 unique video files in the current training split. Intervals range
+from 0.4 to 30 seconds. 861 training rows have no accepted cue and use the original
+prompt as their teacher input. Validation has no cue annotations.
+
+## Scope and limitations
+
+- `current` excludes four confirmed unreadable video intervals and eight training
+  records tied to seven confirmed overlapping source videos. This is not an
+  exhaustive disjointness guarantee against every benchmark.
+- An existing audit identified 252 current MCQs without an explicit option block.
+  Their mapping is unresolved; this release preserves them and publishes stable
+  IDs in `metadata/known_issues.json`.
+- Cues are generated annotations. Automated screening is not a guarantee that a
+  cue is correct or free of answer-bearing information.
+- Historical cue generation could inspect the full source clip, including content
+  after a query timestamp. These are **not certified prefix-only cues**.
+- Student input intervals end at the recorded query timestamp. This does not
+  certify that every answer or cue is recoverable from the sampled student frames.
+- Training interval sampling and four-frame streaming deployment are distinct.
+  No rendered clips or precomputed frame cache are distributed here.
+- This dataset release does not establish new benchmark scores or independent
+  training-seed uncertainty.
+
+## Provenance, terms and integrity
+
+See [video sources](docs/video_sources.md), [construction and preprocessing](docs/data_processing.md),
+and [component terms](LICENSE.md). ThinkStream's MIT notice is included under `licenses/`.
+Source videos remain subject to their original distribution conditions; the dataset
+repository does not grant rights to those videos.
+
+`metadata/release_manifest.json` identifies the release files and upstream revisions.
+`checksums.sha256` enables local integrity checks. `metadata/video_manifest.parquet`
+lists unique source video paths without internal machine paths.
+
+If you use these annotations, acknowledge StreamOPD-Data and the upstream ThinkStream,
+LLaVA-Video and Tarsier2/Kinetics sources. Use the repository revision when reporting
+the precise data version; no unverified paper citation is fabricated here.

checksums.sha256

@@ -0,0 +1,20 @@
+010cbd2f10f80ff0de2b7f6fe9b2a6f60bf16f4bf1636ad840c74ce5d5e1f33e  LICENSE.md
+0a3f33180cbc07773416abddc85b80bb7754551a5b8f9da5f665d7951b7966d3  README.md
+6887f88259f1b8e54be28133464b08d4ca037eefed93a93bfb62614e74592a72  data/current/train.parquet
+ec779eaa241f4b55e919191a0c895afab8d18ffa179cf3ac72d5a170bef4ef6d  data/current/validation.parquet
+a63e3035d9c7a499d8c826cef0c59117ac21c0f0e3824da73800385a9c87364d  data/grpo_25114/train.parquet
+88299ed2c6cac7d821ddacf9e78ffde45011044173a727bd0f1a498d3f9ca441  data/grpo_25114/validation.parquet
+ee9428d82454ca32e6372150f65c092afb2c00b0e8c068de20f9bef3c809ebfd  data/legacy_25118/train.parquet
+ec779eaa241f4b55e919191a0c895afab8d18ffa179cf3ac72d5a170bef4ef6d  data/legacy_25118/validation.parquet
+bea35f3f796a92e6ba1bb89519a8dbb9c6781a8cd2e6ea80d42d98380135fe16  docs/data_processing.md
+2802ec00968864629dd976d1bd799cc678ebbb7be5d0808fd3a7ee0ac5e04d34  docs/video_sources.md
+1a6e572118d667a39799ae3d06234531192dc35d50f331de11738946fabe48ac  licenses/ThinkStream-NOTICE.md
+5a32a327f87e2111bc7f5120545585de9594b2d367d387711d52a146306977e7  metadata/dataset_stats.json
+8a1c5f768d1654362fa922f220eaedf2e9a8dbe469837530f9bffa8a4c68db9c  metadata/exclusions.json
+e6eab5e249d0bf5c5ca1c9b74917498dc8def936867549db884c5e816a8cd9aa  metadata/known_issues.json
+44714b82157cc746a877a47cff1c59f35b3f16a61c0e7a4f3b5d9b3cae6c8218  metadata/release_manifest.json
+0b73117ae5d7713a21d29b9bde5604f97ac04a0ecfa33260a85a7d7275083845  metadata/upstream_sources.json
+56a43fa7adfcd3da8a21bfea87e1b2ea126e4fa8e6d3d603d8b1dddf15138a65  metadata/video_manifest.parquet
+05c8291308a5cc40a78f8b2d7875476945185c8ed11ebfe479288a8744d6a2cc  requirements.txt
+544c9dfa62b0aa485eee7c9329eb5b394426deea46d9f8be079210cfaee1338b  scripts/prepare_local_data.py
+f3ef61676994602a0668dda91cf22b262935bff961c43a096c7b6b8518343046  scripts/validate_data.py

data/current/train.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6887f88259f1b8e54be28133464b08d4ca037eefed93a93bfb62614e74592a72
+size 12945036

data/current/validation.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ec779eaa241f4b55e919191a0c895afab8d18ffa179cf3ac72d5a170bef4ef6d
+size 181959

data/grpo_25114/train.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a63e3035d9c7a499d8c826cef0c59117ac21c0f0e3824da73800385a9c87364d
+size 11579117

data/grpo_25114/validation.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:88299ed2c6cac7d821ddacf9e78ffde45011044173a727bd0f1a498d3f9ca441
+size 185399

data/legacy_25118/train.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ee9428d82454ca32e6372150f65c092afb2c00b0e8c068de20f9bef3c809ebfd
+size 12951276

data/legacy_25118/validation.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ec779eaa241f4b55e919191a0c895afab8d18ffa179cf3ac72d5a170bef4ef6d
+size 181959

docs/data_processing.md

@@ -0,0 +1,111 @@
+# Construction, prompts and temporal processing
+
+## Annotation lineage
+
+The source QA annotations come from ThinkStream's `streaming_rlvr_processed.jsonl`
+and `streaming_cot_cold_processed_5_20.jsonl`. The construction pipeline selected
+verifiable multiple-choice, binary and counting answers and converted them into
+single-question chat records with source-video intervals.
+
+The earlier base pool combined an 8,343-row pass-rate-selected subset and a
+23,884-row verifiable subset, deduplicating on question text and ground truth to
+produce 25,118 rows. The historical pass-rate selector used eight samples per
+student/teacher and retained teacher score >= student score while excluding cases
+where both were perfect. The published artifacts are the retained pool; complete
+historical generation and sampling traces are not included.
+
+`legacy_25118` stores that 25,118-row pool together with its explicit cue-instruction
+teacher prompt. Ignoring the teacher fields yields the corresponding plain OPD QA
+inputs. This configuration intentionally retains four previously confirmed
+unreadable intervals and the eight known overlapping-source records.
+
+`grpo_25114` removes those four unreadable intervals and retains the direct-answer
+prompt suffixes actually used by the GRPO comparison. It does not contain cues and
+does not apply the later eight-row source-overlap exclusion.
+
+`current` additionally removes eight records tied to seven confirmed training/OVO
+source-video overlaps. It stores the original student prompt, cue-positive teacher
+prompt, raw reference, and matched-empty reference. `metadata/exclusions.json`
+records the exact stable IDs and reasons, without internal filesystem information.
+
+## Release transformations
+
+The publication conversion:
+
+1. Replaces internal absolute video paths with source-relative paths.
+2. Renames the original numeric row index to `source_row_id` and adds a split-aware
+   `sample_id`; the local conversion tool restores the trainer's row-index alias.
+3. Gives all configurations one explicit Arrow schema and removes embedded pandas
+   metadata. No binary video data is embedded.
+4. Reconstructs convenience reference fields for historical configurations. It does
+   not claim these references were used in historical training.
+5. Replaces email text in two questions' answer choices with `[REDACTED_EMAIL]` in
+   all corresponding prompts. The source IDs are 9390 and 20692; answer letters and
+   ordering are unchanged. `privacy_redacted` marks those rows.
+
+Except for these two documented contact-text replacements, original questions,
+answers, cue text, prompt wrappers, video interval values and row order are preserved.
+No credentials, machine addresses, private author names, internal logs or source
+absolute paths are intentionally distributed.
+
+## Video intervals and frames
+
+`video_start` and `video_end` are seconds relative to the referenced video file.
+The conversion uses a maximum preceding window of 30 seconds:
+
+```text
+start = max(0, query_timestamp - 30)
+end   = query_timestamp
+```
+
+A timestamp audit matched all 25,106 current rows to source query timestamps and
+found their `video_end` equal to the query time. 445 training rows have nonzero start
+times. Stored intervals are authoritative for reconstruction; do not replace them
+with whole-video sampling.
+
+The verified training input preflight used decord on CPU, one decoder thread, and
+qwen-vl-utils 0.0.14 frame-range and frame-count helpers. Frame indices were generated
+by uniformly spacing between the decoder-resolved start and end indices and rounding
+to integer frame indices. The four unreadable intervals failed this input check.
+
+The release does not bake a universal FPS, number of frames or pixel budget into
+the dataset. Actual frame counts and resizing depend on the model processor and
+resolved training configuration. Keep those settings with experiment records.
+Qwen3.5 and Qwen3-VL use different processors and model modes. For CueGate, the
+teacher positive/reference scoring branches reuse the student's decoded visual
+payload while changing the text prompt.
+
+Four recent frames at deployment are a separate evaluation protocol. Do not
+reinterpret every training example as exactly four frames. Frame-index sampling
+alone is not a certification of strict absolute presentation-timestamp causality.
+
+## Visual cues and reference prompts
+
+Existing cues were generated with Qwen3.5-27B, using instructions to point to a time
+or visual region rather than answer the question. Generation stripped recognized
+answer-option blocks and used automated leakage checks/retries. The historical
+stripping did not cover every option layout, and automated acceptance does not
+prove cue correctness or absence of answer information.
+
+The cue generator passed a source-video URL without applying the student's stored
+start/end interval. Consequently it could inspect the full source clip, including
+post-query content. These existing annotations must not be described as prefix-only.
+A later prefix/full pilot is outside this release; its outputs are not mixed in.
+
+For a row with a cue:
+
+- Student and raw reference: original question and options.
+- Teacher positive: original question plus the cue instruction wrapper and actual cue.
+- Empty reference: same wrapper, with the cue content empty.
+
+For a fallback row, all three teacher message fields equal the original student
+prompt, `cue_available` is false, and `cue_text` is null. Validation has no cues.
+
+## Known data limitations
+
+An existing audit identified 252 current MCQs without an explicit option block.
+The stable IDs are retained in `metadata/known_issues.json`; this publication does
+not invent missing choices or claim to resolve the answer mapping. Known source
+overlap exclusion is incomplete across benchmarks, and no exhaustive human cue
+leakage filter has been applied. Training/validation video IDs should be inspected
+for any additional split-specific use case.

docs/video_sources.md

@@ -0,0 +1,112 @@
+# Video sources and local reconstruction
+
+This repository publishes annotations only. Download upstream videos into a local
+directory after reviewing the upstream terms and access requirements.
+
+| Component | Upstream repository | Pinned revision inspected for this release |
+|---|---|---|
+| QA annotations | `CASIA-IVA-Lab/ThinkStream` | `94a9d82bc74864ba6b705b52ffa3a2d4bc9733e8` |
+| LLaVA video packages | `lmms-lab/LLaVA-Video-178K` | `6d8c562dc26d70042a0d9704d1cae58c94b89098` |
+| Kinetics-700 video package | `omni-research/Tarsier2-Recap-585K` | `5cffa8342bbd9900fb50fdb7fd1b5f2cd1778ac4` |
+
+The last source is access-gated. Request access on its Hugging Face page and log in
+locally before downloading. Keep access tokens out of scripts and command history.
+Pinned repository revisions document upstream file listings; they are not a claim
+that every local video was byte-hashed against that revision.
+
+## Required LLaVA subsets
+
+The release references these five subsets, with their original names:
+
+```text
+0_30_s_youtube_v0_1
+0_30_s_academic_v0_1
+1_2_m_academic_v0_1
+1_2_m_activitynetqa
+2_3_m_academic_v0_1
+```
+
+`metadata/video_manifest.parquet` is the exact list of required files. Archives may
+contain additional videos. Avoid downloading the entire upstream repository if only
+these annotations are needed.
+
+Example, sequential download of the required archive groups:
+
+```python
+from huggingface_hub import snapshot_download
+
+subsets = [
+    "0_30_s_youtube_v0_1", "0_30_s_academic_v0_1",
+    "1_2_m_academic_v0_1", "1_2_m_activitynetqa", "2_3_m_academic_v0_1",
+]
+snapshot_download(
+    repo_id="lmms-lab/LLaVA-Video-178K", repo_type="dataset",
+    revision="6d8c562dc26d70042a0d9704d1cae58c94b89098",
+    allow_patterns=[f"{subset}/*.tar.gz" for subset in subsets],
+    local_dir="downloads/LLaVA-Video-178K", max_workers=1,
+)
+```
+
+Inspect each archive's member paths, then extract each archive under its matching
+subset directory. For example, for one archive:
+
+```bash
+mkdir -p /path/to/videos/LLaVA-Video-178K/0_30_s_youtube_v0_1
+tar -xzf downloads/LLaVA-Video-178K/0_30_s_youtube_v0_1/0_30_s_youtube_v0_1_videos_1.tar.gz \
+  -C /path/to/videos/LLaVA-Video-178K/0_30_s_youtube_v0_1
+```
+
+Historical annotations contain `.../<subset>/data/<source-folder>/...`. The archive
+may instead extract `.../<subset>/<source-folder>/...`. The preparation script tries
+both forms, so a machine-specific symlink is unnecessary. `liwei_youtube_videos`,
+`academic_source`, and `ActivityNet-QA` are inherited upstream directory names.
+
+## Kinetics-700 package
+
+```python
+from huggingface_hub import snapshot_download
+
+snapshot_download(
+    repo_id="omni-research/Tarsier2-Recap-585K", repo_type="dataset",
+    revision="5cffa8342bbd9900fb50fdb7fd1b5f2cd1778ac4",
+    allow_patterns=["Kinetics-700/videos.tar.part-*", "Kinetics-700/metadata.json"],
+    local_dir="downloads/Tarsier2", max_workers=1,
+)
+```
+
+The inspected package has 13 numbered `videos.tar.part-*` files. After obtaining all
+parts, concatenate in filename order and extract the tar stream:
+
+```bash
+mkdir -p /path/to/videos/Kinetics-700
+cat downloads/Tarsier2/Kinetics-700/videos.tar.part-* \
+  | tar -xf - -C /path/to/videos/Kinetics-700
+```
+
+The archive contains `videos/<clip-id>.mp4`. Clip filenames such as
+`<video-id>_000179_000189.mp4` identify an upstream clip; the annotation interval is
+relative to that clip, not the original full YouTube timeline. Re-downloading from
+another source can change clip boundaries or availability, so verify correspondence.
+
+## Expected layout
+
+```text
+videos/
+├── LLaVA-Video-178K/
+│   ├── 0_30_s_youtube_v0_1/liwei_youtube_videos/...
+│   ├── 0_30_s_academic_v0_1/academic_source/...
+│   ├── 1_2_m_academic_v0_1/academic_source/...
+│   ├── 1_2_m_activitynetqa/ActivityNet-QA/...
+│   └── 2_3_m_academic_v0_1/academic_source/...
+└── Kinetics-700/videos/...
+```
+
+Validate path mapping without writing output:
+
+```bash
+python scripts/prepare_local_data.py --video-root /path/to/videos --config current --check-only
+```
+
+Use `--llava-root` and `--kinetics-root` if sources are stored separately. The check
+does not download, decode or modify videos. File existence alone cannot guarantee
+correct clip identity, decodability, or the validity of every time interval.

licenses/ThinkStream-NOTICE.md

@@ -0,0 +1,42 @@
+# ThinkStream source notice
+
+Source dataset: https://huggingface.co/datasets/CASIA-IVA-Lab/ThinkStream
+
+Inspected revision: `94a9d82bc74864ba6b705b52ffa3a2d4bc9733e8`.
+The source dataset card declares `license: mit` and provides this citation:
+
+```bibtex
+@misc{liu2026thinkingstreamingvideo,
+  title={Thinking in Streaming Video},
+  author={Zikang Liu and Longteng Guo and Handong Li and Ru Zhen and Xingjian He and Ruyi Ji and Xiaoming Ren and Yanhao Zhang and Haonan Lu and Jing Liu},
+  year={2026},
+  eprint={2603.12938},
+  archivePrefix={arXiv},
+  primaryClass={cs.CV},
+  url={https://arxiv.org/abs/2603.12938}
+}
+```
+
+The public upstream authors listed here are source attribution, not private
+publisher profile information. The source card did not include a separate
+copyright-holder line; this release does not invent one.
+
+## MIT permission and warranty text
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in
+all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+THE SOFTWARE.

metadata/dataset_stats.json

@@ -0,0 +1,104 @@
+{
+  "current": {
+    "train": {
+      "rows": 25106,
+      "cue_available": 24245,
+      "privacy_redacted_rows": 2,
+      "response_formats": {
+        "Multiple Choice": 17485,
+        "Counting": 1257,
+        "Binary": 6364
+      },
+      "video_source_rows": {
+        "Kinetics-700": 5292,
+        "LLaVA-Video-178K": 19814
+      },
+      "unique_videos": 22848,
+      "parquet_bytes": 12945036
+    },
+    "validation": {
+      "rows": 326,
+      "cue_available": 0,
+      "privacy_redacted_rows": 0,
+      "response_formats": {
+        "Multiple Choice": 259,
+        "Binary": 48,
+        "Counting": 19
+      },
+      "video_source_rows": {
+        "Kinetics-700": 64,
+        "LLaVA-Video-178K": 262
+      },
+      "unique_videos": 326,
+      "parquet_bytes": 181959
+    }
+  },
+  "legacy_25118": {
+    "train": {
+      "rows": 25118,
+      "cue_available": 24257,
+      "privacy_redacted_rows": 2,
+      "response_formats": {
+        "Multiple Choice": 17496,
+        "Counting": 1258,
+        "Binary": 6364
+      },
+      "video_source_rows": {
+        "Kinetics-700": 5292,
+        "LLaVA-Video-178K": 19826
+      },
+      "unique_videos": 22859,
+      "parquet_bytes": 12951276
+    },
+    "validation": {
+      "rows": 326,
+      "cue_available": 0,
+      "privacy_redacted_rows": 0,
+      "response_formats": {
+        "Multiple Choice": 259,
+        "Binary": 48,
+        "Counting": 19
+      },
+      "video_source_rows": {
+        "Kinetics-700": 64,
+        "LLaVA-Video-178K": 262
+      },
+      "unique_videos": 326,
+      "parquet_bytes": 181959
+    }
+  },
+  "grpo_25114": {
+    "train": {
+      "rows": 25114,
+      "cue_available": 0,
+      "privacy_redacted_rows": 2,
+      "response_formats": {
+        "Multiple Choice": 17493,
+        "Counting": 1257,
+        "Binary": 6364
+      },
+      "video_source_rows": {
+        "Kinetics-700": 5292,
+        "LLaVA-Video-178K": 19822
+      },
+      "unique_videos": 22855,
+      "parquet_bytes": 11579117
+    },
+    "validation": {
+      "rows": 326,
+      "cue_available": 0,
+      "privacy_redacted_rows": 0,
+      "response_formats": {
+        "Multiple Choice": 259,
+        "Binary": 48,
+        "Counting": 19
+      },
+      "video_source_rows": {
+        "Kinetics-700": 64,
+        "LLaVA-Video-178K": 262
+      },
+      "unique_videos": 326,
+      "parquet_bytes": 185399
+    }
+  }
+}

metadata/exclusions.json

@@ -0,0 +1,102 @@
+[
+  {
+    "sample_id": "streamopd-train-000454",
+    "source_row_id": 454,
+    "excluded_from": [
+      "current"
+    ],
+    "reason": "confirmed_training_evaluation_source_video_overlap"
+  },
+  {
+    "sample_id": "streamopd-train-001537",
+    "source_row_id": 1537,
+    "excluded_from": [
+      "current"
+    ],
+    "reason": "confirmed_training_evaluation_source_video_overlap"
+  },
+  {
+    "sample_id": "streamopd-train-006663",
+    "source_row_id": 6663,
+    "excluded_from": [
+      "current"
+    ],
+    "reason": "confirmed_training_evaluation_source_video_overlap"
+  },
+  {
+    "sample_id": "streamopd-train-011567",
+    "source_row_id": 11567,
+    "excluded_from": [
+      "current"
+    ],
+    "reason": "confirmed_training_evaluation_source_video_overlap"
+  },
+  {
+    "sample_id": "streamopd-train-013397",
+    "source_row_id": 13397,
+    "excluded_from": [
+      "current"
+    ],
+    "reason": "confirmed_training_evaluation_source_video_overlap"
+  },
+  {
+    "sample_id": "streamopd-train-013520",
+    "source_row_id": 13520,
+    "excluded_from": [
+      "current",
+      "grpo_25114"
+    ],
+    "reason": "confirmed_unreadable_video_interval"
+  },
+  {
+    "sample_id": "streamopd-train-014573",
+    "source_row_id": 14573,
+    "excluded_from": [
+      "current"
+    ],
+    "reason": "confirmed_training_evaluation_source_video_overlap"
+  },
+  {
+    "sample_id": "streamopd-train-015620",
+    "source_row_id": 15620,
+    "excluded_from": [
+      "current"
+    ],
+    "reason": "confirmed_training_evaluation_source_video_overlap"
+  },
+  {
+    "sample_id": "streamopd-train-018971",
+    "source_row_id": 18971,
+    "excluded_from": [
+      "current"
+    ],
+    "reason": "confirmed_training_evaluation_source_video_overlap"
+  },
+  {
+    "sample_id": "streamopd-train-022633",
+    "source_row_id": 22633,
+    "excluded_from": [
+      "current",
+      "grpo_25114"
+    ],
+    "reason": "confirmed_unreadable_video_interval"
+  },
+  {
+    "sample_id": "streamopd-train-023628",
+    "source_row_id": 23628,
+    "excluded_from": [
+      "current",
+      "grpo_25114"
+    ],
+    "reason": "confirmed_unreadable_video_interval"
+  },
+  {
+    "sample_id": "streamopd-train-024965",
+    "source_row_id": 24965,
+    "excluded_from": [
+      "current",
+      "grpo_25114"
+    ],
+    "reason": "confirmed_unreadable_video_interval"
+  }
+]

metadata/known_issues.json

@@ -0,0 +1,1013 @@
+{
+  "scope": "Existing audit of current training rows; annotations unchanged.",
+  "records": [
+    {
+      "sample_id": "streamopd-train-001315",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-001328",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-001345",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-001386",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-001411",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-001450",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-001577",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-001614",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-001705",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-001713",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-001751",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-001775",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-001782",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-001921",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-001956",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-001989",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-002023",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-002127",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-002212",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-002456",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-002521",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-002588",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-002618",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-002653",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-002665",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-002780",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-002834",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-002857",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-002892",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-002961",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-003015",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-003053",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-003095",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-003149",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-003194",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-003204",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-003246",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-003253",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-003256",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-003279",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-003291",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-003336",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-003343",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-003432",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-003546",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-003565",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-003581",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-003582",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-003669",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-004035",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-004057",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-004063",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-004171",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-004241",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-004259",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-004340",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-004389",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-004410",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-004436",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-004465",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-004489",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-004516",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-004606",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-004640",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-004756",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-004776",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-004810",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-004885",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-004909",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-004982",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-005003",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-005064",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-005106",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-005255",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-005289",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-005332",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-005343",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-005413",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-005415",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-005483",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-005515",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-005532",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-005574",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-005598",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-005646",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-005658",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-005730",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-005749",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-005762",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-005815",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-005891",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-006008",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-006019",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-006047",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-006054",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-006087",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-006093",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-006187",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-006218",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-006270",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-006327",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-006436",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-006455",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-006457",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-006580",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-006611",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-006645",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-006649",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-006667",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-006693",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-006702",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-006711",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-006768",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-006830",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-006875",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-006896",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-006922",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-006923",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-007093",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-007131",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-007169",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-007236",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-007270",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-007333",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-007335",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-007432",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-007503",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-007521",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-007524",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-007554",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-007566",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-007661",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-007683",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-007910",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-007992",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-008152",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-008166",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-008199",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-008305",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-008453",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-008589",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-009026",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-009070",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-009141",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-009365",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-009446",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-009453",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-009575",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-009590",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-010122",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-010517",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-010985",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-011043",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-011064",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-011277",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-011304",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-011572",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-011612",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-011708",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-011751",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-011949",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-012050",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-012247",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-012543",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-012691",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-012697",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-012723",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-012892",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-013015",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-013040",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-013046",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-013051",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-013438",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-013514",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-013562",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-013681",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-013840",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-013894",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-014079",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-014101",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-014191",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-014352",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-014560",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-014749",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-014798",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-015409",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-015547",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-015595",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-015701",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-015705",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-016169",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-016447",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-016823",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-016862",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-016903",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-017041",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-017075",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-017447",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-017467",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-017490",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-017630",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-017672",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-018059",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-018606",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-018715",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-018889",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-019368",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-019422",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-019551",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-019915",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-020556",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-020584",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-020602",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-020850",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-020998",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-021108",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-021200",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-021430",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-021626",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-021675",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-022006",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-022104",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-022156",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-022382",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-022416",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-022459",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-022504",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-022526",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-022641",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-022650",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-022698",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-022700",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-022707",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-022915",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-022932",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-022960",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-023315",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-023761",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-023895",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-024038",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-024120",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-024207",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-024225",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-024229",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-024320",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-024375",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-024458",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-024588",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-024669",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-024847",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-025074",
+      "issue": "mcq_without_explicit_option_block"
+    },
+    {
+      "sample_id": "streamopd-train-025110",
+      "issue": "mcq_without_explicit_option_block"
+    }
+  ]
+}

metadata/release_manifest.json

@@ -0,0 +1,153 @@
+{
+  "release_id": "v1",
+  "created_date": "2026-11-02",
+  "dataset_repo": "UniX-Lab/StreamOPD-Data",
+  "videos_included": false,
+  "default_config": "current",
+  "input_fingerprints": {
+    "current/train": {
+      "sha256": "15aa5d0fedd879daa2fa428cb81880f9444421ba9ce7adf67aa5b8e35711cbc2",
+      "rows": 25106
+    },
+    "current/validation": {
+      "sha256": "1fb9b70c687e39d365ec509b3052834298cc7e584bd2f9b1dcda97853dfdc21f",
+      "rows": 326
+    },
+    "legacy_25118/train": {
+      "sha256": "82d056476e38676f25ee4ed7f514bbdff062cb47acf79f4cfdaebe98714476d2",
+      "rows": 25118
+    },
+    "legacy_25118/validation": {
+      "sha256": "2e57c88fd4670fe378bb2288a9d0c53a0af7e579778cdcbe486e2d2f1cdac860",
+      "rows": 326
+    },
+    "grpo_25114/train": {
+      "sha256": "a83c355cbcd18cc2fbee51d939ba1732dd84ecf1ba85c5a08151a101b2090f58",
+      "rows": 25114
+    },
+    "grpo_25114/validation": {
+      "sha256": "da60c7a37ff000c5933fddaf7ea06fe9546c954d0fb67de90156538f387ca7f8",
+      "rows": 326
+    }
+  },
+  "transforms": [
+    "absolute_to_source_relative_video_paths",
+    "uniform_explicit_arrow_schema",
+    "remove_embedded_schema_metadata",
+    "stable_split_aware_ids",
+    "email_text_redaction_in_two_source_questions",
+    "convenience_teacher_reference_fields"
+  ],
+  "validation": {
+    "all_six_splits_roundtrip_verified": true,
+    "original_content_preserved_except_documented_email_redactions": true,
+    "release_privacy_pattern_hits": 0,
+    "local_video_path_mapping": {
+      "config": "current",
+      "train_rows": 25106,
+      "validation_rows": 326,
+      "unique_video_paths": 23174,
+      "resolved_unique_video_paths": 23174,
+      "missing_references": 0,
+      "annotation_and_interval_roundtrip": "passed",
+      "video_decoding_performed": false
+    }
+  },
+  "files": [
+    {
+      "path": "LICENSE.md",
+      "bytes": 1287,
+      "sha256": "010cbd2f10f80ff0de2b7f6fe9b2a6f60bf16f4bf1636ad840c74ce5d5e1f33e"
+    },
+    {
+      "path": "README.md",
+      "bytes": 8291,
+      "sha256": "0a3f33180cbc07773416abddc85b80bb7754551a5b8f9da5f665d7951b7966d3"
+    },
+    {
+      "path": "data/current/train.parquet",
+      "bytes": 12945036,
+      "sha256": "6887f88259f1b8e54be28133464b08d4ca037eefed93a93bfb62614e74592a72"
+    },
+    {
+      "path": "data/current/validation.parquet",
+      "bytes": 181959,
+      "sha256": "ec779eaa241f4b55e919191a0c895afab8d18ffa179cf3ac72d5a170bef4ef6d"
+    },
+    {
+      "path": "data/grpo_25114/train.parquet",
+      "bytes": 11579117,
+      "sha256": "a63e3035d9c7a499d8c826cef0c59117ac21c0f0e3824da73800385a9c87364d"
+    },
+    {
+      "path": "data/grpo_25114/validation.parquet",
+      "bytes": 185399,
+      "sha256": "88299ed2c6cac7d821ddacf9e78ffde45011044173a727bd0f1a498d3f9ca441"
+    },
+    {
+      "path": "data/legacy_25118/train.parquet",
+      "bytes": 12951276,
+      "sha256": "ee9428d82454ca32e6372150f65c092afb2c00b0e8c068de20f9bef3c809ebfd"
+    },
+    {
+      "path": "data/legacy_25118/validation.parquet",
+      "bytes": 181959,
+      "sha256": "ec779eaa241f4b55e919191a0c895afab8d18ffa179cf3ac72d5a170bef4ef6d"
+    },
+    {
+      "path": "docs/data_processing.md",
+      "bytes": 5961,
+      "sha256": "bea35f3f796a92e6ba1bb89519a8dbb9c6781a8cd2e6ea80d42d98380135fe16"
+    },
+    {
+      "path": "docs/video_sources.md",
+      "bytes": 4323,
+      "sha256": "2802ec00968864629dd976d1bd799cc678ebbb7be5d0808fd3a7ee0ac5e04d34"
+    },
+    {
+      "path": "licenses/ThinkStream-NOTICE.md",
+      "bytes": 1890,
+      "sha256": "1a6e572118d667a39799ae3d06234531192dc35d50f331de11738946fabe48ac"
+    },
+    {
+      "path": "metadata/dataset_stats.json",
+      "bytes": 2365,
+      "sha256": "5a32a327f87e2111bc7f5120545585de9594b2d367d387711d52a146306977e7"
+    },
+    {
+      "path": "metadata/exclusions.json",
+      "bytes": 2335,
+      "sha256": "8a1c5f768d1654362fa922f220eaedf2e9a8dbe469837530f9bffa8a4c68db9c"
+    },
+    {
+      "path": "metadata/known_issues.json",
+      "bytes": 27568,
+      "sha256": "e6eab5e249d0bf5c5ca1c9b74917498dc8def936867549db884c5e816a8cd9aa"
+    },
+    {
+      "path": "metadata/upstream_sources.json",
+      "bytes": 507,
+      "sha256": "0b73117ae5d7713a21d29b9bde5604f97ac04a0ecfa33260a85a7d7275083845"
+    },
+    {
+      "path": "metadata/video_manifest.parquet",
+      "bytes": 499078,
+      "sha256": "56a43fa7adfcd3da8a21bfea87e1b2ea126e4fa8e6d3d603d8b1dddf15138a65"
+    },
+    {
+      "path": "requirements.txt",
+      "bytes": 49,
+      "sha256": "05c8291308a5cc40a78f8b2d7875476945185c8ed11ebfe479288a8744d6a2cc"
+    },
+    {
+      "path": "scripts/prepare_local_data.py",
+      "bytes": 5520,
+      "sha256": "544c9dfa62b0aa485eee7c9329eb5b394426deea46d9f8be079210cfaee1338b"
+    },
+    {
+      "path": "scripts/validate_data.py",
+      "bytes": 2969,
+      "sha256": "f3ef61676994602a0668dda91cf22b262935bff961c43a096c7b6b8518343046"
+    }
+  ]
+}

metadata/upstream_sources.json

@@ -0,0 +1,20 @@
+[
+  {
+    "repo_id": "CASIA-IVA-Lab/ThinkStream",
+    "revision": "94a9d82bc74864ba6b705b52ffa3a2d4bc9733e8",
+    "gated": false,
+    "declared_license": "mit"
+  },
+  {
+    "repo_id": "lmms-lab/LLaVA-Video-178K",
+    "revision": "6d8c562dc26d70042a0d9704d1cae58c94b89098",
+    "gated": false,
+    "declared_license": null
+  },
+  {
+    "repo_id": "omni-research/Tarsier2-Recap-585K",
+    "revision": "5cffa8342bbd9900fb50fdb7fd1b5f2cd1778ac4",
+    "gated": "auto",
+    "declared_license": "apache-2.0"
+  }
+]

metadata/video_manifest.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:56a43fa7adfcd3da8a21bfea87e1b2ea126e4fa8e6d3d603d8b1dddf15138a65
+size 499078

requirements.txt

@@ -0,0 +1,3 @@
+pyarrow>=14
+datasets>=2.19
+huggingface_hub>=0.24

scripts/prepare_local_data.py

@@ -0,0 +1,125 @@
+#!/usr/bin/env python3
+"""Resolve public relative video paths into local, trainer-compatible Parquet.
+
+No download or video decoding is performed. Existing output files are not replaced.
+"""
+import argparse
+import copy
+import json
+from functools import lru_cache
+from pathlib import Path, PurePosixPath
+
+import pyarrow as pa
+import pyarrow.parquet as pq
+
+
+def candidates(relative, video_root, llava_root=None, kinetics_root=None):
+    parts = PurePosixPath(relative).parts
+    if not parts or PurePosixPath(relative).is_absolute() or '..' in parts:
+        raise ValueError('Video paths must be relative without parent traversal')
+    if parts[0] == 'LLaVA-Video-178K':
+        base = llava_root or video_root / 'LLaVA-Video-178K'
+        yield base.joinpath(*parts[1:])
+        # Original video archives often extract source folders directly, without
+        # the historical extra 'data/' alias used in annotation paths.
+        if len(parts) > 3 and parts[2] == 'data':
+            yield base.joinpath(parts[1], *parts[3:])
+    elif parts[0] == 'Kinetics-700':
+        if kinetics_root:
+            yield kinetics_root.joinpath(*parts[1:])
+        else:
+            if directory_exists(video_root / 'Kinetics-700'):
+                yield video_root.joinpath(*parts)
+            if directory_exists(video_root / 'tarsier2_unzip' / 'Kinetics-700'):
+                yield video_root.joinpath('tarsier2_unzip', *parts)
+    else:
+        raise ValueError('Unsupported video source: ' + parts[0])
+
+
+@lru_cache(maxsize=65536)
+def resolved_parent(parent):
+    return parent.resolve()
+
+
+@lru_cache(maxsize=65536)
+def directory_exists(path):
+    return path.is_dir()
+
+
+@lru_cache(maxsize=65536)
+def directory_files(path):
+    # Scandir avoids one remote metadata round-trip per video on shared storage.
+    if not directory_exists(path):
+        return frozenset()
+    import os
+    with os.scandir(path) as entries:
+        return frozenset(e.name for e in entries if e.is_file())
+
+
+def file_exists(path):
+    return path.name in directory_files(path.parent)
+
+
+def resolved_file(path):
+    return path.resolve() if path.is_symlink() else resolved_parent(path.parent) / path.name
+
+
+def main():
+    ap = argparse.ArgumentParser(description=__doc__)
+    ap.add_argument('--release-dir', type=Path, default=Path(__file__).resolve().parents[1])
+    ap.add_argument('--config', choices=['current', 'legacy_25118', 'grpo_25114'], default='current')
+    ap.add_argument('--video-root', type=Path, required=True)
+    ap.add_argument('--llava-root', type=Path, help='Optional extracted LLaVA-Video-178K root')
+    ap.add_argument('--kinetics-root', type=Path, help='Optional Kinetics-700 root containing videos/')
+    ap.add_argument('--output-dir', type=Path)
+    ap.add_argument('--check-only', action='store_true')
+    args = ap.parse_args()
+    if not args.check_only and args.output_dir is None:
+        ap.error('--output-dir is required unless --check-only is set')
+    tables = {}
+    missing = []
+    mappings = {}
+    for split in ['train', 'validation']:
+        table = pq.read_table(args.release_dir / 'data' / args.config / (split + '.parquet'))
+        videos = copy.deepcopy(table['videos'].to_pylist())
+        for row_id, row in enumerate(videos):
+            for video in row:
+                relative = video['video']
+                if relative not in mappings:
+                    matches = {resolved_file(p) for p in candidates(relative, args.video_root, args.llava_root, args.kinetics_root)
+                               if file_exists(p)}
+                    if len(matches) > 1:
+                        raise ValueError('Multiple distinct files match ' + relative)
+                    mappings[relative] = str(next(iter(matches))) if matches else None
+                resolved = mappings[relative]
+                if resolved is None:
+                    missing.append({'split': split, 'row': row_id, 'video': relative})
+                else:
+                    video['video'] = resolved
+            if (row_id + 1) % 5000 == 0:
+                print(json.dumps({'split': split, 'checked_rows': row_id + 1}), flush=True)
+        table = table.set_column(table.schema.get_field_index('videos'), table.schema.field('videos'),
+                                 pa.array(videos, type=table.schema.field('videos').type))
+        # The project trainer uses this numeric ID to join original row records.
+        table = table.append_column('rebuttal_row_id', table['source_row_id'])
+        tables[split] = table
+    summary = {'config': args.config, 'unique_video_paths': len(mappings),
+               'resolved_unique_video_paths': sum(p is not None for p in mappings.values()),
+               'missing_references': len(missing), 'missing_examples': missing[:10],
+               'rows': {s: t.num_rows for s, t in tables.items()}, 'video_decoding_performed': False}
+    print(json.dumps(summary, indent=2))
+    if missing:
+        raise SystemExit('Missing videos; no training files were written')
+    if args.check_only:
+        return
+    args.output_dir.mkdir(parents=True, exist_ok=True)
+    outputs = [args.output_dir / (split + '.parquet') for split in tables]
+    if any(p.exists() for p in outputs):
+        raise SystemExit('Output exists; choose a new output directory')
+    for split, table in tables.items():
+        pq.write_table(table, args.output_dir / (split + '.parquet'), compression='zstd')
+    print('Wrote local train.parquet and validation.parquet')
+
+
+if __name__ == '__main__':
+    main()

scripts/validate_data.py

@@ -0,0 +1,67 @@
+#!/usr/bin/env python3
+"""Validate release checksums, relative paths, schema and split consistency."""
+import argparse
+import hashlib
+import json
+import math
+from pathlib import Path, PurePosixPath
+import re
+
+import pyarrow.parquet as pq
+
+
+def string_values(value):
+    if isinstance(value, str):
+        yield value
+    elif isinstance(value, dict):
+        for v in value.values():
+            yield from string_values(v)
+    elif isinstance(value, list):
+        for v in value:
+            yield from string_values(v)
+
+
+def main():
+    ap = argparse.ArgumentParser(description=__doc__)
+    ap.add_argument('--release-dir', type=Path, default=Path(__file__).resolve().parents[1])
+    args = ap.parse_args()
+    root = args.release_dir
+    checksum_file = root / 'checksums.sha256'
+    if checksum_file.is_file():
+        for line in checksum_file.read_text().splitlines():
+            expected, rel = line.split('  ', 1)
+            assert hashlib.sha256((root / rel).read_bytes()).hexdigest() == expected, rel
+    stats = json.loads((root / 'metadata/dataset_stats.json').read_text())
+    result = []
+    for config, splits in stats.items():
+        seen = set()
+        for split, expected in splits.items():
+            table = pq.read_table(root / 'data' / config / (split + '.parquet'))
+            assert table.schema.metadata is None
+            rows = table.to_pylist()
+            assert len(rows) == expected['rows']
+            assert sum(row['cue_available'] for row in rows) == expected['cue_available']
+            assert sum(row['privacy_redacted'] for row in rows) == expected['privacy_redacted_rows']
+            for row in rows:
+                assert row['sample_id'] not in seen
+                seen.add(row['sample_id'])
+                assert row['ground_truth'] == row['reward_model']['ground_truth']
+                assert row['teacher_reference_raw'] == row['prompt']
+                if not row['cue_available']:
+                    assert row['cue_text'] is None
+                    assert row['teacher_prompt'] == row['prompt'] == row['teacher_reference_empty']
+                for video in row['videos']:
+                    p = PurePosixPath(video['video'])
+                    assert not p.is_absolute() and '..' not in p.parts
+                    assert p.parts[0] in ['LLaVA-Video-178K', 'Kinetics-700']
+                    assert math.isfinite(video['video_start']) and math.isfinite(video['video_end'])
+                    assert 0 <= video['video_start'] < video['video_end']
+                for value in string_values(row):
+                    assert not re.search(r'(?<![\w-])hf_[A-Za-z0-9]{20,}', value)
+                    assert not re.search(r'[\w.%+-]+@[\w.-]+\.[A-Za-z]{2,}', value)
+            result.append({'config': config, 'split': split, 'rows': len(rows), 'status': 'passed'})
+    print(json.dumps({'checksums_verified': checksum_file.is_file(), 'splits': result}, indent=2))
+
+
+if __name__ == '__main__':
+    main()