Tag
This paper quantifies the compression cost of pre-tokenisation boundary rules by bounding minimum token counts from both sides, certifying that regex boundaries increase optimal token counts by 28.3–36.8% on English Wikipedia, and showing that compression-optimal token dictionaries do not necessarily improve language-model prediction quality.
Anthropic has released 13 free AI courses with certificates, covering topics from introduction to advanced skills like API usage and model context protocol, aimed at practical AI education.
Introduces 'Training Under Challenge', an executable-certificate framework that uses architecture-valid procedures to construct alternative models and estimate the empirical global-optimality gap of neural network checkpoints, with theoretical guarantees and experiments on ResNet-18 distillation and quantized denoising.
This paper presents a certificate-carrying sub-quadratic method for computing bisimulation metrics in Markov decision processes using approximate nearest neighbors, with coverage-augmented guarantees and two-sided bounds. Experiments show improved scaling and accurate metric recovery compared to baselines.
This paper introduces the notion of publicly-verifiable certificates of statistical validity (pvCSVs) for statistical algorithms, enabling distributionally-robust certification of learning results without interaction. The authors construct pvCSVs for adaptive Statistical Query algorithms with sample complexity scaling logarithmically in the number of queries.
Explains how to set up TLS certificates for internal services using split-horizon DNS, a VPN with DNS resolver, and ACME clients like acme.sh with Let's Encrypt, providing a practical alternative to self-signed certificates.
The article discusses the problem of authentication token theft by infostealer malware and explores a 15-year-old proposal by Dirk Balfanz to use self-signed client certificates for TLS mutual authentication to bind tokens to a specific device, preventing token reuse even if stolen.
This article alerts Linux distributions about the upcoming expiration of Microsoft's UEFI CA certificates used for Secure Boot, detailing new certificates and potential boot issues on newer hardware that lacks the old ones.
This paper formalizes hallucination-to-action conversion in multimodal agents and proposes evidence-carrying agents (ECA) that use constrained verifiers to authorize only safe tool calls, achieving 0% unsafe-action rate on a 200-task pipeline.