AWS added optional prompt shortening to Bedrock Advanced Prompt Optimization. AWS says the service returns an optimized prompt only when it is shorter and scores at least as high under the configured steering criteria, model judge or Lambda evaluator, and reports token counts plus a compression ratio.
Why it matters: The feature makes prompt length an optimization objective alongside task quality, which can reduce input cost and time to first token. Its usefulness still depends on whether the evaluation actually represents production behavior.
Google Research highlighted WikiProfile, a 2,150-fact benchmark spanning ten tasks and thirteen models, in an analysis of whether models encode facts they fail to answer directly. The authors report behavioral evidence that leading models encode 95–98% of tested facts but miss 26–34% under direct recall; extended reasoning reduces the failure rate to 11–12% and recovers 40–65% of encoded-but-not-directly-known facts.
Why it matters: If many factual failures are retrieval failures rather than absent knowledge, evaluation and system design should separately measure encoding, direct recall and recovery under reasoning or prompting.