Research map / Learning theory and algorithms / Prediction and decision losses
Losses for decision-making
Design losses for learning to defer, predict-then-optimize and risk criteria beyond average error.
68 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Prediction and decision losses, keyword cluster “losses · loss functions · surrogate”
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.
22 papers in 2023–24 and 13 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 4.1% to 2.3%.
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 · Minimizing the Maximal Loss: How and Why (ICML 2016 · 35 citations)
- 2018 · A Batch Learning Framework for Scalable Personalized Ranking (AAAI 2018 · 2 citations)
- 2019 · Melding the Data-Decisions Pipeline: Decision-Focused Learning for Combinatorial Optimization (AAAI 2019 · 214 citations)
- 2020 · Consistent Estimators for Learning to Defer to an Expert (ICML 2020 · 24 citations)
- 2021 · GLISTER: Generalization based Data Subset Selection for Efficient and Robust Learning (AAAI 2021 · 125 citations)
- 2022 · Decision-Focused Learning: Through the Lens of Learning to Rank (ICML 2022 · 16 citations)
- 2023 · UltraRE: Enhancing RecEraser for Recommendation Unlearning via Error Decomposition (NeurIPS 2023 · 4 citations)
- 2024 · Model Alignment as Prospect Theoretic Optimization (ICML 2024 · 23 citations)
- 2025 · ALTBI: Constructing Improved Outlier Detection Models via Optimization of Inlier-Memorization Effect (AAAI 2025 · 0 citations)
- 2026 · Generalizing Linear Autoencoder Recommenders with Decoupled Expected Quadratic Loss (ICLR 2026 · 0 citations)
Earliest papers here
- Minimizing the Maximal Loss: How and Why (ICML 2016 · 35 citations)
- A Batch Learning Framework for Scalable Personalized Ranking (AAAI 2018 · 2 citations)
- Convex Elicitation of Continuous Properties (NeurIPS 2018 · 1 citation)
What has been tried
56 papers here report 518 measured changes (a component added, removed or swapped, compared with a baseline): helped 63%, no clear effect 12%, hurt 19%, mixed 6%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| use one output per input instead of both preference outputs | GSM8K (8-shot, chain-of-thought) · EM | 53.5 -> 50.0 | Hurt | Model Alignment as Prospect Theoretic Optimization (ICML 2024 · 23 citations) |
| Add feature-based parameterization (FBP) to LODL | Cubic Top-K · Normalized Test Decision Quality | -0.38 ± 0.29 -> 0.58 ± 0.26 | Helped | Leaving the Nest: Going beyond Local Loss Functions for Predict-Then-Optimize (AAAI 2024 · 11 citations) |
| fix the reparameterization matrices to the identity | Quadratic OU (easy) · control L2 error | Hurt | Stochastic Optimal Control Matching (NeurIPS 2024 · 1 citation) | |
| Use adaptive loss truncation and incrementally increasing mini-batches | ADBench datasets · AUC | Helped | ALTBI: Constructing Improved Outlier Detection Models via Optimization of Inlier-Memorization Effect (AAAI 2025 · 0 citations) | |
| Assign shared margin parameters to minority classes and distinct parameters to frequent classes | CIFAR-10, CIFAR-100, and Tiny ImageNet step-imbalanced settings · accuracy | No clear effect | Balancing the Scales: A Theoretical and Algorithmic Framework for Learning from Imbalanced Data (ICML 2025 · 0 citations) | |
| Use exit layer 12 instead of layer 1 for loss-based SIFT | BERT validation set, last step · validation accuracy | 0.6599 -> 0.6640 | Helped | Understanding the Training Speedup from Sampling with Approximate Losses (ICML 2024 · 0 citations) |
| Vary the smoothing parameter | Inference tasks across model architectures · performance | No clear effect | Coded Computing for Resilient Distributed Computing: A Learning-Theoretic Framework (NeurIPS 2024) |
What reviewers will expect
Datasets and baselines named in the 11 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- CIFAR-10 · 4 of 11 papers
- CIFAR-100 · 3 of 11 papers
- SVHN · 2 of 11 papers
- Tiny ImageNet · 2 of 11 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 35 papers; from OpenAlex and the papers' affiliations).
- Stanford University · 6 papers
- Google · 4 papers
- Courant Institute of Mathematical Sciences · 4 papers
- Cornell University · 3 papers
- Nanyang Technological University · 3 papers
- New York University · 3 papers
- Apple · 3 papers
- Harvard University · 2 papers
- Princeton University · 2 papers
- Columbia University · 2 papers
Use it
Download BibTeX of 38 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
- Consistency of surrogate losses (72 papers)
- Loss landscape analysis (61 papers)
- Ranking-based loss optimization (50 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)
- Ranking models from comparisons (127 papers)
- Determinantal point process inference (122 papers)
- Conformal prediction sets (116 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).