Research map / Label-efficient and robust learning / Recognition and long-tail
Parallel-mentoring for Offline Model-based Optimization
Can Chen, Christopher Beckham, Zixuan Liu, Xue Liu, Christopher Pal
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add voting-based pairwise supervision (vs mean-ensemble labels) (ablation) | tri-mentoring · Design-Bench NAS | max normalized score: 0.509 -> 0.759 | Improved |
| add voting-based pairwise supervision (vs mean-ensemble labels) (ablation) | tri-mentoring · Design-Bench Hopper Controller | max normalized score: 1.902 -> 1.983 | Improved |
| add voting-based pairwise supervision (vs mean-ensemble labels) (ablation) | tri-mentoring · Design-Bench TF Bind 8 | max normalized score: 0.971 -> 0.970 | No clear effect |
| add adaptive soft-labeling (vs one-hot consensus labels) (ablation) | tri-mentoring · Design-Bench NAS | max normalized score: 0.538 -> 0.759 | Improved |
| add adaptive soft-labeling (vs one-hot consensus labels) (ablation) | tri-mentoring · Design-Bench TF Bind 8 | max normalized score: 0.944 -> 0.970 | Improved |
| add neighborhood sampling around the optimization point (ablation) | tri-mentoring · Design-Bench Hopper Controller | max normalized score: 1.839 -> 1.983 | Improved |
| add neighborhood sampling around the optimization point (ablation) | tri-mentoring · Design-Bench Superconductor | max normalized score: 0.516 -> 0.514 | No clear effect |
| proxies exchange ranking signals instead of post selection (ablation) | tri-mentoring · Design-Bench TF Bind 10 | max normalized score: 0.635 -> 0.722 | Improved |
| proxies exchange ranking signals instead of post selection (ablation) | tri-mentoring · Design-Bench D'Kitty Morphology | max normalized score: 0.970 -> 0.966 | No clear effect |
| add voting-based supervision on label accuracy of conflicting labels (ablation) | tri-mentoring · D'Kitty pairwise labels | accuracy of conflicting labels (%): 52.21% -> 62.91% | Improved |
| add adaptive soft-labeling on label accuracy of conflicting labels (ablation) | tri-mentoring · TFB8 pairwise labels | accuracy of conflicting labels (%): 60.86% -> 63.16% | Improved |
| combine COMs with tri-mentoring | tri-mentoring · Ant Morphology / TF Bind 8 | max normalized score: 0.948 -> 0.941 (Ant), 0.970 -> 0.960 (TFB8) | Hurt |
Closest papers by meaning
- Robust Model-Based Optimization for Challenging Fitness Landscapes (ICLR 2024 · similarity 0.86)
- Offline Model-Based Optimization via Policy-Guided Gradient Search (AAAI 2024 · similarity 0.86)
- Learning Surrogates for Offline Black-Box Optimization via Gradient Matching (ICML 2024 · similarity 0.85)
- Bridging Model-Based Optimization and Generative Modeling via Conservative Fine-Tuning of Diffusion Models (NeurIPS 2024 · similarity 0.85)
- Training Greedy Policy for Proposal Batch Selection in Expensive Multi-Objective Combinatorial Optimization (ICML 2024 · similarity 0.85)
- ROOT: Rethinking Offline Optimization as Distributional Translation via Probabilistic Bridge (NeurIPS 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
