Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
The Label Imitation Game: Turing Test Network for Zero-Shot Pseudo-Label Pruning
Brent Griffin, Jason Corso
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 |
|---|---|---|---|
| prune pseudo-labels with zero-shot TTN | VLM pseudo-labels (AL) · Overall pseudo-labels, 4 datasets | F1: 0.362 -> 0.376 | Improved |
| fine-tune TTN on detection data (TTN_D) | TTN · Overall pseudo-labels, 4 datasets | F1: 0.376 -> 0.387 | Improved |
| TTN pruning on min-F1 (hardest) classes | AL · Min-F1 classes mean | F1: 0.133 -> 0.170 | Improved |
| TTN_D pruning on min-F1 classes | AL · Min-F1 classes mean | F1: 0.133 -> 0.192 | Improved |
| prune pseudo-labels (zero-shot TTN) on LVIS | AL · LVIS pseudo-labels | F1: 0.078 -> 0.079 | No clear effect |
| train detector on TTN-pruned labels | YOLO11n on AL labels · Overall downstream | mAP50: 0.350 -> 0.355 | Improved |
| train detector on TTN_D-pruned labels | YOLO11n on AL labels · Overall vulnerable classes | mAP50: 0.155 -> 0.171 | Improved |
| pruning reduces recall | AL · Overall pseudo-labels | recall: 0.453 -> 0.437 | Hurt |
| use CLIP kNN relabeling instead of pruning | AL · Pseudo-label evaluation | F1: 0.362 -> 0.275 | Hurt |
| use TTN tokenizer embeddings for kNN instead of CLIP (ablation) | kNN prune · Pseudo-label evaluation | F1: 0.366 -> 0.373 | Improved |
| use ground-truth reference set instead of self-referential (ablation) | TTN · Pseudo-label evaluation | F2: 0.386 -> 0.395 | Improved |
| shrink self-referential reference set to <=100 (ablation) | TTN · Pseudo-label evaluation | F2: 0.386 -> 0.381 | Hurt |
Closest papers by meaning
- All Labels Are Not Created Equal: Enhancing Semi-Supervision via Label Grouping and Co-Training (CVPR 2021 · similarity 0.87)
- Label-Efficient Dataset Pruning via Semi-Supervised Pseudo-Labeling (NeurIPS 2026 · similarity 0.86)
- Non-Stationary Predictions May Be More Informative: Exploring Pseudo-Labels with a Two-Phase Pattern of Training Dynamics (ICML 2025 · similarity 0.86)
- Pre-Trained Vision-Language Models as Noisy Partial Annotators (AAAI 2025 · similarity 0.86)
- Iterative Self-Training with Class-Aware Text-to-Image Synthesis for Visual Task Learning (AAAI 2025 · similarity 0.86)
- Efficient and Context-Aware Label Propagation for Zero-/Few-Shot Training-Free Adaptation of Vision-Language Model (ICLR 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.
