Research map / Learning theory and algorithms / Prediction and decision losses
Ranking models from comparisons
Model and infer rankings and preferences from pairwise or partial comparisons, e.g. Mallows models.
127 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Prediction and decision losses, keyword cluster “ranking · nearest · neighbor”
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.
24 papers in 2023–24 and 26 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic held steady, from 4.5% to 4.5%.
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 · A Comparative Study of Ranking-Based Semantics for Abstract Argumentation (AAAI 2016 · 100 citations)
- 2017 · Preference Elicitation For Participatory Budgeting (AAAI 2017 · 82 citations)
- 2018 · To Trust Or Not To Trust A Classifier (NeurIPS 2018 · 278 citations)
- 2019 · Joint Optimization of Tree-based Index and Deep Model for Recommender Systems (NeurIPS 2019 · 31 citations)
- 2020 · Fast Differentiable Sorting and Ranking (ICML 2020 · 11 citations)
- 2021 · Towards Efficient Selection of Activity Trajectories based on Diversity and Coverage (AAAI 2021 · 31 citations)
- 2022 · Obtaining Calibrated Probabilities with Personalized Ranking Models (AAAI 2022 · 19 citations)
- 2023 · Uniform Sequence Better: Time Interval Aware Data Augmentation for Sequential Recommendation (AAAI 2023 · 92 citations)
- 2024 · Interactive Hyperparameter Optimization in Multi-Objective Problems via Preference Learning (AAAI 2024 · 5 citations)
- 2025 · Foundations of the Theory of Performance-Based Ranking (CVPR 2025 · 3 citations)
- 2026 · Predictable Compression Failures: Order Sensitivity and Information Budgeting for Evidence-Grounded Binary Adjudication (ICML 2026 · 1 citation)
Earliest papers here
- A Comparative Study of Ranking-Based Semantics for Abstract Argumentation (AAAI 2016 · 100 citations)
- Deconvolving Feedback Loops in Recommender Systems (NeurIPS 2016 · 43 citations)
- Data-driven Rank Breaking for Efficient Rank Aggregation (ICML 2016 · 21 citations)
What has been tried
104 papers here report 870 measured changes (a component added, removed or swapped, compared with a baseline): helped 64%, no clear effect 11%, hurt 21%, mixed 5%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Replace linear RankSVM with non-linear RankSVM | LCBench 32 datasets · end-to-end HPO performance | No clear effect | Interactive Hyperparameter Optimization in Multi-Objective Problems via Preference Learning (AAAI 2024 · 5 citations) | |
| Add per-item parameters to the contextual model | ImageClarity, WiscAds, and IMDB-WIKI-SbS · holdout comparison error | Helped | Active preference learning for ordering items in- and out-of-sample (NeurIPS 2024 · 5 citations) | |
| Replace Sibling ranking with Outshines ranking | Delphi daily deployment · maximally tied outliers | 14k ± 1.7k -> 21 ± 5.5 | Helped | Outlier Ranking for Large-Scale Public Health Data (AAAI 2024 · 2 citations) |
| Use multiple anchor selection vectors rather than one | Diginetica · approximation error | 19.2% to 15.2% | Helped | Would You Like Your Data to Be Trained? A User Controllable Recommendation Framework (AAAI 2024 · 1 citation) |
| add L_e2e to LCRON | RecFlow Period 1 last-day test, two-stage cascade · Joint Recall@10@20 | 0.8710 ± 0.0013 -> 0.8732 ± 0.0005 | Helped | Learning Cascade Ranking as One Network (ICML 2025 · 0 citations) |
| Include tied pairs in standard DPO training | NMT and TL;DR · task performance | Hurt | On Extending Direct Preference Optimization to Accommodate Ties (NeurIPS 2025 · 0 citations) | |
| Add YouTube channel embeddings | Tournesol YouTube comparisons, validation (10-fold CV) · validation risk | 10.1 \cdot 10^{-3} -> 8.40 \cdot 10^{-3}; 17% reduction | Helped | Generalizing while preserving monotonicity in comparison-based preference learning models (NeurIPS 2025 · 0 citations) |
| Use backtracking during noisy insertion search instead of naive insertion without backtracking | Synthetic seriation scenarios 1–4 · probability of error | Helped | Active Seriation: Efficient Ordering Recovery with Statistical Guarantees (NeurIPS 2025 · 0 citations) | |
| Increase user-random-choice probability ζ | MovieLens10M and ACSEmployment · overall loss | Hurt | Learning from Streaming Data when Users Choose (ICML 2024 · 0 citations) | |
| larger subset sizes and wider inter-alternative gaps | MTurk pilot · ground-truth recovery | Helped | The Surprising Effectiveness of SP Voting with Partial Preferences (NeurIPS 2024 · 0 citations) |
What reviewers will expect
Datasets and baselines named in the 16 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Baselines
- DPO · 3 of 16 papers
- IPO · 3 of 16 papers
- Borda · 2 of 16 papers
- RLHF · 2 of 16 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 50 papers; from OpenAlex and the papers' affiliations).
- Institute of Logic and Computation · 2 papers
- Alibaba Group · 2 papers
- Calicarpa · 2 papers
- Huawei · 2 papers
- Télécom ParisTech · 2 papers
- Technion - Israel Institute of Technology · 2 papers
- Cornell University · 2 papers
- Amazon · 2 papers
- Google · 2 papers
Use it
Download BibTeX of 40 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
- Approximate nearest neighbor indexing (58 papers)
- Metric and ordinal embedding (47 papers)
- Locality-sensitive hashing (30 papers)
Elsewhere in the topic
- Adaptive testing and experiments (201 papers)
- Active learning and subset selection (199 papers)
- Distributionally robust optimization (184 papers)
- Data attribution and recourse (160 papers)
- Theory of neural network sensitivity (143 papers)
- Determinantal point process inference (122 papers)
- Conformal prediction sets (116 papers)
- Provable post-hoc explanations (92 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).