AI Briefing#test-time-scaling
The latest AI and developer news about #test-time-scaling, with the original source and a short summary.
The latest AI and developer news about #test-time-scaling, with the original source and a short summary.
The paper distinguishes standard test-time scaling from compute extrapolation, where performance continues to improve at test-time compute levels far exceeding the maximum used during training.
LLM-as-a-Verifier v0.2.0 has been released, achieving SOTA performance on agent benchmarks such as Terminal-Bench 2.1 and reducing uncached input tokens by approximately 3.4x through cache optimization.
Proposes the CLR framework, which verifies errors at the claim level to improve test-time inference efficiency.
Proposes a risk control framework that optimizes token usage during LLM inference while keeping the error rate below a certain level.
AutoTTS reduced LLM inference tokens by 70% through AI agent-based iterative optimization.
AutoTTS discovers Test-Time Scaling policies through code exploration alone in an offline replay environment.
Combining summarization, voting, and re-refinement boosted the performance of long-horizon coding agents.