The Archive

Finite-width signal propagation analysis shows when infinite-width approximation breaks down in long linear recurrences.

Proposes Prefix Sampling to optimize RL training efficiency by maintaining 50% pass rate—the regime maximizing reward signal and entropy in agentic tasks like SWE-bench.

LineRides framework enables bicycle robot to learn complex stunts via line-guided RL without demonstrations, using spatial guidelines and sparse keyframe constraints.

Framework for materials science dataset construction balancing targeted property optimization against preservation of untargeted outcomes via diversity-aware selection.

Introduces Concept Field method to detect hallucination and measure novelty in LLM outputs by modeling semantic drift in text corpora using sentence embeddings.

Unified theoretical framework for distributional regret bounds in bandits and episodic RL, with UCBVI-style algorithm achieving gap-independent guarantees.

Anthropic partners with SpaceX for compute capacity; removes Claude Code peak-hour limits and raises API rate limits for Opus.

Analysis of 100 most popular hardware configurations for local LLM inference on Hugging Face reveals deployment patterns and infrastructure preferences.

Memini: associative memory system with multi-timescale dynamics for continual knowledge updating in deployed LLMs without explicit management.

Bayesian framework for active view selection in 3D reconstruction using posterior inference over implicit surfaces.

Doubly sparse regularization exploiting Gaussian graphical model structure for high-dimensional regression.

Announced in the code w/claude programme. Hope to see claude back on top especially with the new rate limits!

Driver-WM: latent world model for predicting driver reactions during L2/L3 automation transitions using in-cabin behavioral dynamics.

Think-aloud traces improve automated cognitive model discovery beyond behavior-only constraints in risky decision-making tasks.

Reddit speculation about Claude usage limit increases allegedly tied to SpaceX partnership; unverified claim with screenshot evidence.

https://x.ai/news/anthropic-compute-partnership

Automated pipeline for auditing unexpected behavioral side-effects of LLM interventions through contrastive multi-token generation analysis.

User reports Claude's rate limit counter reset unexpectedly from 99% to 0%.

Link: x.com

Anthropic partners with SpaceX for compute capacity; removes Claude Code peak-hour limits and raises API rate limits for Opus.

Gated multimodal model combining EPC tabular data and assessor text to predict building energy efficiency scores.

ORDERED: variance reduction for unsupervised domain adaptation via optimal data reordering during training.

Imitation learning for stabilizing Vlasov-Poisson plasma control using sparse macroscopic diagnostics with stability guarantees.

Psychometric analysis of 50 LLMs identifies phenomenal experience as primary variance axis via Pinocchio dimension.

Also, rate limits will double for Pro and Max users of tools like Claude Code.

I’ve been using Claude mostly for coding and summarizing boring work docs, but today it accidentally became my cyber security therapist. I got an email from what looked exactly like one of my vendors asking me to update payment info for an invoice. Same writing style, same signature, referenced a real project, everything. I was literally about to send the payment when something felt slightly off, but I couldn’t explain why. Out of curiosity I pasted the email into Claude and asked if anything looked suspicious. It immediately pointed out a bunch of manipulation tactics I completely missed, ...

Vision-based mmWave beam management system for V2X vehicular connectivity using camera sensing and closed-loop learning.

Theoretical proof that long-context models cannot simultaneously optimize efficiency, compactness, and recall—fundamental trade-off affecting Transformers and SSMs.

Google demonstrates AI Mode in Search with gardening tips; product feature announcement, not technical or architectural interest.