arXiv:2607. 18451v1 Announce Type: cross Abstract: Frozen encoders are chosen by how well a lightweight head reads a finding from their features, not whether the geometry separates it.
Frozen encoders are chosen by how well a lightweight head reads a finding from their features, not whether the geometry separates it. Nearest-neighbor discordance does, but with unequal banks the opposite-label neighbor wins on density, not geometry, so prevalence alone makes an uninformed encoder look blind.
arXiv:2607.18451v3 Announce Type: replace Abstract: A foundation encoder is pretrained once on a large image corpus and then reused with its weights frozen. Each new task is solved by training a smal...
arXiv:2608. 16709v1 Announce Type: cross Abstract: A radiologist reading a model's output faces two problems.
The paper critiques current memorization audits for generative models, arguing that lacking a proper null distribution leads to misleading conclusions. It introduces two exact null tests—one permutation test for whole models and a calibrated test for single images—showing that many previously flagged memorizations disappear under these stricter controls. The authors also propose a scale‑restricted statistic based on the Intersection Euler Characteristic Profile to better detect distinct copied images.
arXiv:2604.11508v3 Announce Type: replace Abstract: Fine-tuning a pretrained classifier leaves some samples reliably learned and others cycling between correct and incorrect. Curriculum learning, dat...
The paper demonstrates that open‑ended Theory‑of‑Mind trackers can produce valid beliefs that are absent from finite reference sets, and that treating unmatched outputs as false can reverse model‑selection rankings. By recoding references for 259 beliefs, the authors show a dramatic drop in weighted prevalence and a reversal of strictly proper Brier risk, with similar distortions observed in a 301‑question NQ‑open DPR‑BERT pipeline. The study further reveals that 90‑96% of audited unmatched beliefs are literally true, and introduces a TriSource‑Restore method that anchors reference labels to a probability‑sampled human pilot to restore calibration and ranking integrity.
The paper demonstrates that sharing a deep encoder alone does not eliminate the confounding effects in task-comparison scores. By introducing a conditional two‑discriminator discrepancy within the embedding space, the authors achieve robust detection of task changes, maintaining stability under input rotations and accurately tracking label‑permutation drift. This approach, integrated into a mixture‑of‑heads framework, outperforms traditional novelty triggers and generalizes across multiple backbones and datasets, including ImageNet‑21k ViT‑B/16, DINOv2, and CIFAR‑100.
arXiv:2609.35470v1 Announce Type: cross Abstract: Applied researchers increasingly convert images into features with pretrained encoders, then use those features in a downstream prediction model. The...
The paper investigates whether response safety can be measured by the cosine similarity between a response embedding and the mean embedding of known‑safe responses. Using four frozen encoders and prompt‑controlled datasets, the authors find that a simple prototype (mean safe embedding) performs poorly (ROC‑AUC 0.457‑0.545) while an explicit safe‑minus‑unsafe reference achieves higher scores (0.588‑0.738). The study shows that a class mean is merely a location, not a safety direction, and that a reference with sufficient unsafe mass is needed to orient safety judgments.
The paper evaluates nine on‑device named‑entity recognition models ranging from classical taggers to large language models, measuring not only accuracy but also latency and output validity. Using a silver‑gold benchmark derived from an LLM judge panel and a human‑validated corpus, the study shows that encoder‑based models achieve comparable accuracy to a 4 B instruct LLM while being much smaller, faster, and producing no malformed output. Confidence calibration of GLiNER is analyzed, revealing over‑confidence but improved reliability after temperature scaling and thresholding.
arXiv:2608. 15428v1 Announce Type: cross Abstract: Multiple-choice benchmarks are graded on whether a model picks the right option, not on whether it needed the question.
The paper investigates drift detection in deep learning models, showing that sharing a deep encoder alone does not eliminate confounding in task-comparison scores. By introducing a conditional two‑discriminator discrepancy into the embedding space, the authors create a two‑axis gate that remains stable under input rotations and accurately tracks label‑permutation drift. This approach outperforms traditional exchange or novelty triggers, achieving high AUROC in distinguishing semantic novelty from photometric shift across multiple backbones and datasets.