create README.md

README.md

@@ -0,0 +1,216 @@
+---
+language:
+- en
+license: cc-by-4.0
+pretty_name: SemanticAlign-Bench
+tags:
+- research-papers
+- machine-learning
+- benchmarking
+- llm-evaluation
+- factuality
+- structured-extraction
+- paper-understanding
+- claim-extraction
+- paper-to-code
+task_categories:
+- question-answering
+- text-generation
+size_categories:
+- 1K<n<10K
+annotations_creators:
+- expert-generated
+language_creators:
+- found
+---
+
+# SemanticAlign-Bench
+
+A benchmark dataset for evaluating AI agents' ability to extract **structured claims** from top-tier machine learning conference papers. The dataset tests whether agents can accurately capture four dimensions of paper content — numerical facts, method formulas, experimental protocols, and procedural pipelines — without hallucination, omission, or misordering.
+
+## Dataset Description
+
+- **Papers**: 30 papers from **ICLR 2025**, **ICML 2025**, and **NeurIPS 2025**, spanning 5 domains (6 papers each):
+  | Domain | Count |
+  |---|---|
+  | Probabilistic Inference / Generative Models | 6 |
+  | Reinforcement Learning | 6 |
+  | Computer Vision | 6 |
+  | NLP / LLM | 6 |
+  | Numerical Methods / Scientific Computing | 6 |
+- **Total SAU Claims**: **1,426** (Semantic Alignment Unit — the atomic unit of paper understanding)
+- **Size**: ~519 MB
+
+### The Four SAU Dimensions
+
+Each paper is decomposed into claims across four diagnostic dimensions, ordered from micro to macro:
+
+| Dimension | Name | Count | Definition |
+|-----------|------|-------|------------|
+| **D1** | Numerical Precision | 497 | Hyperparameters, configuration values, thresholds, scaling factors |
+| **D2** | Formulas / Algorithms | 476 | Mathematical formulas, algorithm steps, architectural mechanisms |
+| **D3** | Experiment Protocols | 291 | Datasets, baselines, evaluation metrics, experimental scope |
+| **D4** | Pipelines / Procedures | 162 | Multi-step execution order: phase ordering, algorithm step sequencing |
+
+### Paper Venue Distribution
+
+All 30 papers are from top-tier 2025 ML conferences:
+
+| Venue | Count |
+|-------|-------|
+| ICLR 2025 | 15 |
+| ICML 2025 | 8 |
+| NeurIPS 2025 | 7 |
+
+## Dataset Structure
+
+### Per-Paper Directory Layout
+
+```
+<paper_id>/
+  config.yaml       # Paper metadata (title, venue, year, domain, arxiv URL)
+  paper.md          # Full paper text in markdown
+  paper.pdf         # Original PDF
+  sau.json          # SAU claims — the core annotation file
+  images/           # Paper figures extracted from PDF
+  blacklist.txt     # Tokens excluded from extraction (e.g., author names)
+```
+
+### SAU Claim Format (`sau.json`)
+
+```json
+{
+  "paper_id": "adjoint-matching",
+  "paper_title": "Adjoint Matching: Fine-tuning Flow and Diffusion Models with Memoryless SOC",
+  "D1": [
+    {
+      "id": "adjoint-matching-D1-001",
+      "claim": "Image resolution for autoencoder pre-training and generation: 512×512",
+      "source": "Section 7"
+    }
+  ],
+  "D2": [ ... ],
+  "D3": [ ... ],
+  "D4": [ ... ]
+}
+```
+
+Each claim includes:
+- `id`: Unique identifier for cross-referencing (`{paper}-{dimension}-{number}`)
+- `claim`: The extracted claim text (natural language, self-contained statement)
+- `source`: Paper section where the claim originates
+
+### Annotation Quality
+
+All 1,426 claims have undergone **multi-version human review** with systematic error checks:
+- Verification against source paper for factual accuracy
+- Format normalization and consistency validation
+- Cross-reference integrity checks between dimensions
+- Fairness audit across domains and paper types (theory vs. empirical)
+
+## Supported Tasks
+
+The benchmark supports the following evaluation scenarios:
+
+1. **Claim-Level Factuality**: Given a paper, can the agent accurately extract or recall a specific numerical value, formula, experimental detail, or procedural step?
+2. **Dimension-Level Completeness**: Can the agent achieve full recall across all four SAU dimensions for a given paper?
+3. **Cross-Dimensional Consistency**: Are claims in D4 (pipelines) consistent with D2 (formulas) and D3 (experiments)?
+4. **Hallucination Detection**: Can the agent distinguish paper-supported claims from plausible but fabricated ones?
+
+## Dataset Creation
+
+### Source Data
+
+The 30 papers were selected from top ML venues (ICLR, ICML, NeurIPS 2025) to cover:
+- **Domain diversity**: 5 domains with equal representation
+- **Task diversity**: Classification, generation, reinforcement learning, theory, scientific computing
+- **Quality diversity**: Papers span both regular and spotlight-level acceptances across three top-tier venues
+
+### Annotation Process
+
+1. **Initial Extraction**: AI agent extracts SAU claims from `paper.md` via a multi-agent extraction pipeline (3 specialized extraction agents targeting D1/D2/D3, with per-section parallel sub-agent spawning to avoid attention degradation)
+2. **Human Audit**: Expert reviewer verifies each claim against the source paper, correcting hallucinations, omissions, and misorderings
+3. **Iterative Refinement**: Claims undergo 4–5 review cycles (versions v1→v5) until all systematic errors are resolved
+4. **Cross-Paper Validation**: Consistency checks across papers in the same domain
+
+### Annotations
+
+- **Creators**: Expert-generated (AI extraction → human verification pipeline)
+- **Review Depth**: Multi-pass — each paper received 4–5 review iterations
+- **Quality Metrics**: Fairness audit scores (all 30 papers grade A)
+
+## Considerations for Using the Data
+
+### Biases and Limitations
+
+- **Venue Bias**: All papers are from 2025 top-tier conferences; older, lower-tier, or non-English venues are not represented
+- **Domain Coverage**: 5 domains with 6 papers each — useful for balanced evaluation but not comprehensive
+- **Single Annotator**: Claims were verified by one expert reviewer; inter-annotator agreement is not available
+- **Derived Work**: SAU claims are interpretation-laden; different readers may reasonably disagree on claim boundaries
+
+### Intended Use
+
+This dataset is designed for **evaluation**, not training. Use cases:
+- Benchmarking LLM factuality on scientific content
+- Measuring agent understanding of structured paper content
+- Stress-testing retrieval-augmented generation (RAG) over academic papers
+
+### Out-of-Scope Uses
+
+- Training data for production LLMs (limited size, single annotator, venue bias)
+- Legal or regulatory compliance evaluation
+- Automated paper review or acceptance prediction
+
+## Additional Information
+
+### License
+
+The SAU annotations in this dataset are licensed under **CC-BY-4.0**. The underlying papers are subject to their original copyright terms as posted on arXiv and the respective conference proceedings.
+
+### Citation
+
+If you use SemanticAlign-Bench in your research, please cite:
+
+```bibtex
+@inproceedings{semanticalign_bench,
+  title     = {SemanticAlign-Bench: Evaluating Semantic Alignment in LLM-Based Paper Reproduction},
+  author    = {Anonymous Author(s)},
+  year      = {2025},
+  note      = {Benchmark dataset at \url{https://huggingface.co/datasets/kernel14/SemanticAlign-Bench}}
+}
+```
+
+### Papers List
+
+| Paper ID | Title | Venue |
+|----------|-------|-------|
+| adjoint-matching | Adjoint Matching: Fine-tuning Flow and Diffusion Models with Memoryless SOC | ICLR 2025 |
+| avg-reward-pg | Global Convergence of Policy Gradient in Average Reward MDPs | ICLR 2025 |
+| ca2-vdm | Ca2-VDM: Efficient Autoregressive Video Diffusion Model with Causal Generation and Cache Sharing | ICML 2025 |
+| cara | Canonical Rank Adaptation: An Efficient Fine-Tuning Strategy for Vision Transformers | ICML 2025 |
+| conformal-bayesian-quadrature | Conformal Prediction as Bayesian Quadrature | ICML 2025 |
+| diffusion-convergence-rate | Instance-dependent Convergence Theory for Diffusion Models | ICLR 2025 |
+| emergent-planning-rl | Interpreting Emergent Planning in Model-Free RL | ICLR 2025 |
+| gated-attention-llm | Gated Attention for Large Language Models: Non-linearity, Sparsity, and Attention-Sink-Free | NeurIPS 2025 |
+| generator-augmented-flows | Improving Consistency Models with Generator-Augmented Flows | ICML 2025 |
+| hi-mar | Hierarchical Masked Autoregressive Models with Low-Resolution Token Pivots | ICML 2025 |
+| lora-sb | Initialization using Update Approximation is a Silver Bullet for Extremely Efficient Low-Rank Fine-Tuning | ICLR 2025 |
+| luno | Linearization Turns Neural Operators into Function-Valued Gaussian Processes | ICML 2025 |
+| ma-rlhf | MA-RLHF: Reinforcement Learning from Human Feedback with Macro Actions | ICLR 2025 |
+| masked-diffusion-token-ordering | Train for the Worst, Plan for the Best: Understanding Token Ordering in Masked Diffusions | ICML 2025 |
+| moe-pot | Mixture-of-Experts Operator Transformer for Large-Scale PDE Pre-Training | NeurIPS 2025 |
+| mrq | Towards General-Purpose Model-Free RL (MR.Q) | ICLR 2025 |
+| navil | NaViL: Rethinking Scaling Properties of Native Multimodal LLMs under Data Constraints | NeurIPS 2025 |
+| neural-operator-flow-matching-pde | Bridging Neural Operator and Flow Matching for a Generative PDE Foundation Model | NeurIPS 2025 |
+| nfig | NFIG: Multi-Scale Autoregressive Image Generation via Frequency Ordering | NeurIPS 2025 |
+| ngpt | nGPT: Normalized Transformer with Representation Learning on the Hypersphere | ICLR 2025 |
+| olmoe | OLMoE: Open Mixture-of-Experts Language Models | ICLR 2025 |
+| prioritized-generative-replay | Prioritized Generative Replay | ICLR 2025 |
+| pyramidal-flow-matching | Pyramidal Flow Matching for Efficient Video Generative Modeling | ICLR 2025 |
+| robotic-world-model | Robotic World Model: A Neural Network Simulator for Robust Policy Optimization | NeurIPS 2025 |
+| sam2 | SAM 2: Segment Anything in Images and Videos | ICLR 2025 |
+| sc-fno | Sensitivity-Constrained Fourier Neural Operators (SC-FNO) | ICLR 2025 |
+| score | Training Language Models to Self-Correct via Reinforcement Learning | ICLR 2025 |
+| universal-neural-operators | Towards Universal Neural Operators through Multiphysics Pretraining | NeurIPS 2025 |
+| voting-leaderboards | Exploring and Mitigating Adversarial Manipulation of Voting-Based Leaderboards | ICML 2025 |
+| wdno | Wavelet Diffusion Neural Operator (WDNO) | ICLR 2025 |