arXiv:2611.24970 introduces HireWatch, a benchmark for testing whether LLMs select lawful …
PRKV introduces a training-free approach to KV cache eviction. It converts attention relat…
Researchers report 64 from-scratch pretraining runs comparing Transformers, gated linear a…
TETHER turns the table of contents from discarded layout noise into a global structural pr…
The work uses sparse autoencoder features extracted from GraphCast activations to turn a l…
The paper introduces a long-range, video-based face presentation attack detection dataset …
HapBO relocates LLM use out of the Bayesian optimization loop. A human expert, an LLM, or …
This learning-theory preprint claims to settle the weaker form of the sample compression c…
PoolDPP recasts diverse generation as sampling a size-K set of complete strings under a de…
The paper introduces LR-Merge, a method for predicting and correcting interference when se…
The paper challenges a common shortcut: more generated tokens do not monotonically mean th…