Sync accepted camera-ready dataset metadata

Update the dataset card, publication citation, and explicit MIT/CC-BY-4.0 licensing.

CITATION.cff

@@ -0,0 +1,48 @@
+cff-version: 1.2.0
+message: "If you use PAVO-Bench or the meta-controller in your research, please cite it as below."
+title: "PAVO: Pipeline-Aware Voice Orchestration with Demand-Conditioned Inference Routing"
+abstract: >
+  PAVO treats the ASR-LLM-TTS voice pipeline as a jointly optimizable inference graph
+  with demand-conditioned routing. An 85,041-parameter meta-controller, trained with
+  multi-objective PPO in 106 seconds on H100, cuts P95 end-to-end latency by 10.3%
+  on 200 LibriSpeech samples. The corresponding mean-latency reduction is significant
+  at p = 2x10^-6. It reduces per-turn energy by 71% vs a fixed-cloud baseline on a
+  50,000-turn simulated benchmark. Hard-constraint masking on the
+  inter-stage coupling structure (validated across n=5,430 measurements on three
+  LLM families and two hardware platforms) reduces coherence-failure rate from
+  7.1% to 0.9% (7.9x), at +110 ms median latency cost.
+type: software
+license: MIT
+url: "https://github.com/vnmoorthy/pavo-bench"
+repository-code: "https://github.com/vnmoorthy/pavo-bench"
+repository-artifact: "https://huggingface.co/datasets/vnmoorthy/pavo-bench"
+keywords:
+  - voice-pipeline
+  - asr
+  - llm
+  - tts
+  - inference-routing
+  - edge-ai
+  - latency
+  - reinforcement-learning
+  - benchmark
+authors:
+  - family-names: VeiluKanthaPerumal
+    given-names: NarasingaMoorthy
+    affiliation: University of Pennsylvania
+    email: moorthyv@sas.upenn.edu
+  - family-names: Imthathullah
+    given-names: Mohammed
+    affiliation: Google
+preferred-citation:
+  type: article
+  title: "PAVO: Pipeline-Aware Voice Orchestration with Demand-Conditioned Inference Routing"
+  authors:
+    - family-names: VeiluKanthaPerumal
+      given-names: NarasingaMoorthy
+      affiliation: University of Pennsylvania
+    - family-names: Imthathullah
+      given-names: Mohammed
+      affiliation: Google
+  journal: "Transactions on Machine Learning Research"
+  year: 2026

DATA_LICENSE.md

@@ -0,0 +1,11 @@
+# Data, Results, and Model Weights License
+
+Unless a file states otherwise, the datasets, result JSON/JSONL files, coupling
+matrices, generated result figures, and model weights in this repository are
+licensed under the [Creative Commons Attribution 4.0 International License](https://creativecommons.org/licenses/by/4.0/).
+
+Copyright 2026 NarasingaMoorthy VeiluKanthaPerumal and Mohammed Imthathullah.
+
+When reusing these artifacts, provide attribution by citing the PAVO paper or
+the citation in `CITATION.cff`. Source code is separately licensed under the
+[MIT License](LICENSE).

LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 NarasingaMoorthy VeiluKanthaPerumal
+
+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.

README.md

@@ -1,3 +1,7 @@
+<!-- Copy the entire content of this file into the README on
+     https://huggingface.co/datasets/vnmoorthy/pavo-bench (Edit pencil, top right).
+     Preserve the YAML frontmatter; HF requires it for dataset configs. -->
+
 ---
 license: cc-by-4.0
 task_categories:
@@ -24,29 +28,15 @@ size_categories:
 configs:
   - config_name: default
     data_files:
-      - split: tier1_statistical
-        path: tier1_statistical_results.json
-      - split: tier1_coupling
-        path: tier1_coupling_results.json
-      - split: tier1_llm_latency
-        path: tier1_llm_latency_results.json
-      - split: tier2_e2e
-        path: tier2_e2e_results.json
-      - split: tier2_cross_dataset
-        path: tier2_cross_dataset_results.json
-      - split: tier2_noise_robustness
-        path: tier2_noise_robustness_results.json
-      - split: tier3_50k_summary
-        path: tier3_50k_summary.json
-      - split: tier3_scaling
-        path: tier3_scaling_results.json
-      - split: component_ablation
-        path: component_ablation_results.json
+      - split: train
+        path: tier3_50k_train.jsonl
+      - split: test
+        path: tier3_50k_test.jsonl
 ---
 
 # PAVO-Bench: 50K-Turn Benchmark for ASR-LLM-TTS Pipeline Routing
 
-**Code:** [github.com/vnmoorthy/pavo-bench](https://github.com/vnmoorthy/pavo-bench) · **Paper:** TMLR 2026 (under review) · **Authors:** NarasingaMoorthy VeiluKanthaPerumal (UPenn), Mohammed Imthathullah (Google)
+**Code:** [github.com/vnmoorthy/pavo-bench](https://github.com/vnmoorthy/pavo-bench) · **Paper:** [TMLR 2026 (accepted)](https://openreview.net/forum?id=zrneoIxlFx) · **Authors:** NarasingaMoorthy VeiluKanthaPerumal (UPenn), Mohammed Imthathullah (Google)
 
 ```bash
 pip install git+https://github.com/vnmoorthy/pavo-bench.git
@@ -56,12 +46,14 @@ pip install git+https://github.com/vnmoorthy/pavo-bench.git
 
 | Metric | Result | Significance |
 |---|---|---|
-| P95 end-to-end latency (H100, LibriSpeech) | **−10.3%** (−167 ms) | p = 2×10⁻⁶ |
+| P95 end-to-end latency (H100, LibriSpeech) | **−10.3%** (−167 ms) | — |
 | Median latency | **−34%** | |
 | Energy per turn | **−71%** | |
 | Coherence-failure rate | **7.1% → 0.9%** (7.9× reduction) | hard-constraint masking, +110 ms median cost |
 | Meta-controller size | 85,041 parameters | — |
-| Meta-controller training | 106 seconds on A100 | — |
+| Meta-controller training | 106 seconds on H100 | — |
+
+The `p = 2×10⁻⁶` result applies to the paired test of **mean** end-to-end latency (2,277 vs 2,671 ms over five bootstrap replications), not the descriptive P95 comparison; the paired Wilcoxon test on the same five replications gives `p = 0.0625`.
 
 The empirical contribution is a two-regime coupling structure (sharp factual-accuracy cliff + gradual semantic degradation) characterized over **n = 5,430 measurements** across two hardware platforms (H100, Apple M3) and three LLM families (Llama 3.1 8B, Mistral 7B, Gemma2 2B).
 
@@ -69,10 +61,19 @@ The empirical contribution is a two-regime coupling structure (sharp factual-acc
 
 PAVO-Bench evaluates **ASR-LLM-TTS voice pipeline routing** decisions. It provides 50,000 turns of benchmark data designed to measure how well different pipeline configurations balance **latency**, **quality**, **cost**, and **energy** when routing spoken-language queries through cascaded ASR, LLM, and TTS components.
 
-The benchmark is organized into three tiers plus component-level ablation. All results were produced on real GPU hardware.
+The benchmark is organized into three tiers plus component-level ablation. The 50K routing evaluation is simulated using measured H100/M3 latencies and published benchmark values for unavailable configurations; the repository also includes direct-inference H100 results.
 
 ## Dataset Files
 
+The Hugging Face `default` config exposes the 40K/10K JSONL files as standard `train` and `test` splits. The heterogeneous result-summary JSON files below remain directly downloadable artifacts rather than dataset splits, which keeps the Dataset Viewer schema valid.
+
+### Primary dataset
+
+| File | Split | Rows |
+|------|-------|-----:|
+| `tier3_50k_train.jsonl` | train | 40,000 |
+| `tier3_50k_test.jsonl` | test | 10,000 |
+
 ### Tier 1 — Unit-Level Validation
 
 | File | Description |
@@ -125,6 +126,15 @@ pavo  = PretrainedPAVORouter.from_released()
 print(benchmark_router(pavo, turns))
 ```
 
+Or directly with `datasets`:
+
+```python
+from datasets import load_dataset
+
+dataset = load_dataset("vnmoorthy/pavo-bench")
+print(dataset["train"].num_rows, dataset["test"].num_rows)  # 40000 10000
+```
+
 ## Citation
 
 ```bibtex
@@ -138,4 +148,4 @@ print(benchmark_router(pavo, turns))
 
 ## License
 
-CC-BY 4.0
+Dataset, results, coupling matrices, and model weights: CC-BY 4.0. Code is MIT licensed in the linked GitHub repository.