Upload README.md with huggingface_hub

README.md

@@ -1,25 +1,14 @@
 ---
-license: cc-by-nc-4.0
+license: cc-by-4.0
 language:
 - en
-pretty_name: LifeAudioMultiTurnStreamingCoT
 tags:
 - streaming-cot
-- life-scenarios
-- audio
-- multi-turn
-- dialogue
+- chain-of-thought
 - sft
-- reasoning
-- speech
-- timestamped
-task_categories:
-- text-generation
-- question-answering
-- automatic-speech-recognition
-task_ids:
-- language-modeling
-size_categories: 10K<n<100K
+- audio
+size_categories:
+- 1K<n<10K
 configs:
 - config_name: default
   data_files:
@@ -35,111 +24,64 @@ configs:
     path: data/high_quality_eval.parquet
 ---
 
-# LifeAudioMultiTurnStreamingCoT v0.4
-
-**Final professional public release** — clean SFT schema, target field, canonical taxonomy.
-
-## Overview
-
-- **Modality**: Audio
-- **Turn Type**: Multi Turn
-- **Version**: v0.4
-- **License**: cc-by-nc-4.0
-- **Language**: English
-- **HF Repo**: `skyzhou06/LifeAudioMultiTurnStreamingCoT`
+# LifeAudioMultiTurnStreamingCoT
 
-## Row Counts
+**Version:** vFinal
 
-| Split | Rows |
-|-------|------|
-| Train | 10,684 |
-| Eval  | 2,671 |
-| High Quality Train | 10,588 |
-| High Quality Eval  | 2,640 |
-| **Total** | **13,355** |
+A multi-turn audio dataset with streaming chain-of-thought reasoning for SFT. 13349 active training rows with audio references.
 
-## Schema (v0.4)
+## Schema
 
-### Top-Level Fields
+Top-level fields: `id, split, modality, turn_type, dialogue, audio, input, streaming, target, taxonomy, quality, source, metadata`
 
-| Field | Type | Description |
-|-------|------|-------------|
-| `id` | string | Stable example ID |
-| `split` | string | `train` or `eval` |
-| `modality` | string | `audio` |
-| `turn_type` | string | `multi_turn` |
-| `input` | object | Input data |
-| `streaming` | object | Checkpoints with natural-language reasoning |
-| `target` | object | Training target: reasoning, answer, response |
-| `taxonomy` | object | Canonical content classification |
-| `quality` | object | Quality assessment |
-| `source` | object | Provenance |
-| `metadata` | object | Release metadata |
-
-### `target`
+- dialogue (list of turns with role/text)
+- audio (path, reference, duration_sec, loadability)
+- input (instruction, length_bucket)
+- target (reasoning, answer, response)
+- streaming (checkpoints with streaming_reasoning)
+- taxonomy (category, subcategory, difficulty, intent_type)
 
+### Target Format
 ```json
 {
-  "reasoning": "Natural-language reasoning summary.",
-  "answer": "The final answer.",
+  "reasoning": "Natural language reasoning about the task/input...",
+  "answer": "The actual task output...",
   "response": "Reasoning: ...\n\nAnswer: ..."
 }
 ```
 
-- `target.answer` — for answer-only SFT
-- `target.response` — for reasoning-augmented SFT
+Use `target.response` for SFT training. It includes both reasoning and final answer.
 
-### `taxonomy`
+## Quality
 
-Uses canonical enum categories: `daily_life`, `travel`, `education`, `work_productivity`, `finance_consumer`, `health_wellness_safe`, `tech_support`, `information_extraction`, `creative_planning`, `social_communication`.
+| Metric | Value |
+|--------|-------|
+| Active rows | 13,349 |
+| Train | 10,679 |
+| Eval | 2,670 |
+| High quality | 13,349 |
+| SFT-ready | 100.0% |
+| Target grounded | 100.0% |
 
-### `quality`
+## High-Quality Configuration
 
-```json
-{
-  "is_high_quality": true,
-  "sft_ready": true,
-  "natural_reasoning": true,
-  "reasoning_quality": "high",
-  "taxonomy_quality": "high"
-}
-```
+The `high_quality` config contains a filtered subset of default rows where `quality.sft_ready = true` and `quality.is_high_quality = true`. It is not additional unique data.
 
-## Changes from v0.3
+## Limitations
 
-- `output` renamed to `target` with `reasoning`, `answer`, `response`
-- `streaming.trace` removed from active rows
-- `metadata.legacy` blobs removed
-- Taxonomy mapped to canonical enum with fixed category/subcategory pairs
-- `high_quality` split into `high_quality_train` / `high_quality_eval`
-- `metadata.release_version` = `"v0.4"`
+- Audio is external reference only (HF URLs). No bundled audio bytes.
+- 
+- Natural-language reasoning is template-generated, not LLM-written.
+- Row counts reflect quality-filtered active splits suitable for direct SFT usage.
 
-## SFT Usage
+## Usage
 
 ```python
 from datasets import load_dataset
 
-# Default config
+# Load default config
 ds = load_dataset("skyzhou06/LifeAudioMultiTurnStreamingCoT")
-# ds["train"]["target"]["response"]  — reasoning + answer
-# ds["train"]["target"]["answer"]    — answer only
 
-# High quality config
+# Load high-quality subset
 ds_hq = load_dataset("skyzhou06/LifeAudioMultiTurnStreamingCoT", "high_quality")
 ```
-
-## Source Licenses
-
-Dataset-level license: **cc-by-nc-4.0**. Individual rows include `source.license` and `source.dataset` fields with source-specific license information.
-
-**Audio availability**: Most audio is externally referenced (not bundled). See `audio.loadability` field per row.
-
-## Limitations
-
-- Reasoning is rule-based (content-grounded)
-- Some answers are brief closing phrases (check `quality.sft_ready`)
-- Non-English examples not included
-
-## Citation
-
-If you use this dataset, please cite the original source datasets and this release.