Research map / Large language models / AI and society
Benchmarking and reproducibility practice
Promote common task frameworks, dataset documentation, and replication for scientific ML.
104 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in AI and society, keyword cluster “human · systems · argue”
State History Tried Checklist Preprints Who Export
Where it stands
Not enough benchmark data Fewer than four papers report the same benchmark, so there is no record to follow.
38 papers in 2023–24 and 47 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 16.8% to 9.2%.
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2016 · PEAK: Pyramid Evaluation via Automated Knowledge Extraction (AAAI 2016 · 51 citations)
- 2017 · Add Data into Business Process Verification: Bridging the Gap between Theory and Practice (AAAI 2017 · 45 citations)
- 2018 · Toward Driving Scene Understanding: A Dataset for Learning Driver Behavior and Causal Reasoning (CVPR 2018 · 323 citations)
- 2019 · Unreproducible Research is Reproducible (ICML 2019 · 55 citations)
- 2020 · Atari-HEAD: Atari Human Eye-Tracking and Demonstration Dataset (AAAI 2020 · 43 citations)
- 2021 · Therapeutics Data Commons: Machine Learning Datasets and Tasks for Drug Discovery and Development (NeurIPS 2021 · 162 citations)
- 2022 · GeoEngine: A Platform for Production-Ready Geospatial Research (CVPR 2022 · 11 citations)
- 2023 · Toward a Perspectivist Turn in Ground Truthing for Predictive Computing (AAAI 2023 · 87 citations)
- 2024 · Position: Data Authenticity, Consent, & Provenance for AI are all broken: what will it take to fix them? (ICML 2024 · 21 citations)
- 2025 · TabArena: A Living Benchmark for Machine Learning on Tabular Data (NeurIPS 2025 · 11 citations)
- 2026 · No One Knows the State-of-the-Art in Geospatial Foundation Models (NeurIPS 2026 · 2 citations)
Earliest papers here
- PEAK: Pyramid Evaluation via Automated Knowledge Extraction (AAAI 2016 · 51 citations)
- Add Data into Business Process Verification: Bridging the Gap between Theory and Practice (AAAI 2017 · 45 citations)
- A Declarative Approach to Data-Driven Fact Checking (AAAI 2017 · 16 citations)
What has been tried
48 papers here report 342 measured changes (a component added, removed or swapped, compared with a baseline): helped 59%, no clear effect 9%, hurt 25%, mixed 7%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| use holdout validation instead of cross-validation for model selection | TabArena-v0.1 · predictive performance | Hurt | TabArena: A Living Benchmark for Machine Learning on Tabular Data (NeurIPS 2025 · 11 citations) | |
| Omit the e0+ reaction from the abstraction | example with input x:0 and candidate output y:2 · candidate output preservation | Hurt | Shield Synthesis for LTL Modulo Theories (AAAI 2025 · 4 citations) | |
| use forward initialization rather than zero or Gaussian initialization | FashionMNIST · accuracy | Helped | Benchmarking Predictive Coding Networks -- Made Simple (ICLR 2025 · 1 citation) | |
| add causal masking | CarbonSense test sites, excluding WAT · NSE truncated mean | 0.654 -> 0.669 | Helped | CarbonSense: A Multimodal Dataset and Baseline for Carbon Flux Modelling (ICLR 2025 · 0 citations) |
| add point cloud features to visual models | ContPhy physical reasoning settings · accuracy | 52.7 -> 55.0; 53.3 -> 57.7; 67.5 -> 68.3; 45.5 -> 53.4 | Helped | ContPhy: Continuum Physical Concept Learning and Reasoning from Videos (ICML 2024 · 0 citations) |
| Use list-based chunk hashing instead of naive hashing | Files larger than 1 MB · hashing runtime | Helped | Position: Machine Learning Models Have a Supply Chain Problem (ICML 2025) |
What reviewers will expect
Datasets and baselines named in the 37 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- MNIST · 4 of 37 papers
- CIFAR-10 · 3 of 37 papers
Baselines
- XGBoost · 3 of 37 papers
- RandomForest · 2 of 37 papers
- TabDPT · 2 of 37 papers
- LSTM · 2 of 37 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 85 papers; from OpenAlex and the papers' affiliations).
- Google · 6 papers
- Microsoft · 5 papers
- Stanford University · 5 papers
- Inria · 4 papers
- Tsinghua University · 4 papers
- University of Illinois, Urbana Champaign · 4 papers
- Massachusetts Institute of Technology · 4 papers
- University of Oxford · 4 papers
- University of Cambridge · 4 papers
- IBM · 3 papers
Use it
Download BibTeX of 36 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Human-centered trust and agency (161 papers)
- Social-science view of AI evaluation (159 papers)
- Data commons and human baselines (159 papers)
- Critiques of LLM claims and judging (104 papers)
- AI for health and public decisions (55 papers)
Elsewhere in the topic
- Teacher needs and AI curricula (45 papers)
- Classroom tools and student perceptions (40 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).