https://aiunderstanding.org/news/new-benchmark-finds-ai-agents-wrongly-block-approved-work-28-of-the-time AI Understanding en 2026-08-15T06:13:05.761Z New Benchmark Finds AI Agents Wrongly Block Approved Work 28% of the Time https://aiunderstanding.org/news/paper-reports-frontier-llm-judges-flip-verdicts-25-71-under-pushback AI Understanding en 2026-08-15T06:05:35.053Z Paper Reports Frontier LLM Judges Flip Verdicts 25-71% Under Pushback https://aiunderstanding.org/news/paper-argues-evolution-strategies-beat-rl-at-keeping-llm-answer-sets-diverse AI Understanding en 2026-08-15T05:41:29.686Z Paper Argues Evolution Strategies Beat RL at Keeping LLM Answer Sets Diverse https://aiunderstanding.org/news/paper-says-self-improving-ai-agents-can-turn-one-unsafe-success-into-a-reusable-sk AI Understanding en 2026-08-15T05:36:55.310Z Paper Says Self-Improving AI Agents Can Turn One Unsafe Success Into a Reusable Skill https://aiunderstanding.org/news/seag-paper-proposes-aliasing-sensitive-entities-before-rag-queries-reach-external AI Understanding en 2026-08-15T05:30:02.704Z SEAG Paper Proposes Aliasing Sensitive Entities Before RAG Queries Reach External LLMs https://aiunderstanding.org/news/cabs-paper-reports-cheaper-faster-model-merging-across-27-datasets AI Understanding en 2026-08-15T04:35:48.673Z CABS+ Paper Reports Cheaper, Faster Model Merging Across 27 Datasets https://aiunderstanding.org/news/paper-proposes-retrieved-lessons-to-improve-spatial-reasoning-in-frozen-vision-lan AI Understanding en 2026-08-15T04:24:02.431Z Paper Proposes Retrieved "Lessons" to Improve Spatial Reasoning in Frozen Vision-Language Models https://aiunderstanding.org/news/prove-rt-paper-reports-44-7-success-generating-machine-checked-real-time-proofs AI Understanding en 2026-08-15T04:21:05.122Z PROVE-RT Paper Reports 44.7% Success Generating Machine-Checked Real-Time Proofs https://aiunderstanding.org/news/working-paper-asks-whether-india-s-consumer-law-can-cover-ai-harms AI Understanding en 2026-08-15T04:03:29.752Z Working Paper Asks Whether India's Consumer Law Can Cover AI Harms https://aiunderstanding.org/news/apple-paper-proposes-cheaper-machine-unlearning-by-skipping-low-influence-data AI Understanding en 2026-08-13T16:40:29.586Z Apple Paper Proposes Cheaper Machine Unlearning by Skipping Low-Influence Data