Back to reviews
min readarXiv:2603.04731v1
When Priors Backfire: On the Vulnerability of Unlearnable Examples to Pretraining
Authors: Zhihao Li, Gezheng Xu, Jiale Cai, Ruiyi Fang, Di Wu
Pending (κ=0.55)Intermediaterepresentationspuriousalignment
RSCT Score Breakdown
Relevance (R)
0.38
Superfluous (S)
0.32
Noise (N)
0.31
TL;DR
When Priors Backfire: On the Vulnerability of Unlearnable Examples to Pretraining...
When Priors Backfire: On the Vulnerability of Unlearnable Examples to Pretraining
RSCT Certification: κ=0.550 (certified) | RSN: 0.38/0.32/0.31 | Topics: representation, spurious, alignment
Overview
This paper addresses topics relevant to RSCT research, specifically in the areas of representation, spurious, alignment.
Key RSCT Relevance:
- Topic similarity score: 40%
- RSCT whitepaper similarity: 32%
- Combined relevance: 35%
RSCT Quality Metrics
| Metric | Value | Interpretation | |--------|-------|----------------| | κ-gate | 0.550 | Certified | | R (Relevance) | 0.375 | Direct relevance to research goals | | S (Stability) | 0.318 | Supporting context and patterns | | N (Noise) | 0.307 | Irrelevant components | | Gate Reached | 4 | Certification depth |
Paper Details
- Authors: Zhihao Li, Gezheng Xu, Jiale Cai, Ruiyi Fang, Di Wu
- Source: arxiv
- Primary Topic: representation
- Difficulty: Intermediate
This review was auto-generated by the Swarm-It RSCT discovery pipeline.