{"slug": "merit-finds-memory-lifts-tool-agent-success-from-0-00-to-0-55-1-00", "title": "MERIT finds memory lifts tool-agent success from 0.00 to 0.55-1.00", "summary": "A new study on arXiv (2609.05441) finds that adding long-term memory to tool-using large language model agents can lift task success rates from 0.00 to 0.55-1.00, but the choice of memory implementation can shift success by up to 60 points and affect costs by a factor of 2.7-3.9x. The research shows structured fact stores outperform embedding retrieval on updated facts (70-100% success vs. 30-95%), and full replay is never cost-effective, emphasizing the need for careful memory architecture selection in production systems.", "body_md": "[arXiv](https://arxiv.org/abs/2609.05441)\n\n### MERIT finds memory lifts tool-agent success from 0.00 to 0.55-1.00\n\nWhich summary reads better? Pick one — models revealed after.Both summaries are AI-generated.\n\nMemory implementations in tool-using LLM agents can shift task success rates by up to 60 points, with structured fact stores outperforming embedding retrieval on updated facts (70-100% success vs. 30-95%). This means choosing the wrong memory strategy can drastically reduce agent performance, and full replay is never cost-effective—optimal memory use delivers 2.7-3.9x more utility per dollar. Engineers must carefully select memory architectures to maximize both task success and cost efficiency in production systems.\n\nThe marginal utility of long-term memory in tool-using LLM agents can increase task success rates from 0.00 to 0.55-1.00, but the choice of memory implementation can move task success by up to 60 points and affect costs by a factor of 2.7-3.9x; this variability directly impacts the cost-effectiveness and reliability of production LLM agents.", "url": "https://wpnews.pro/news/merit-finds-memory-lifts-tool-agent-success-from-0-00-to-0-55-1-00", "canonical_source": "https://www.snipvote.com/story/cmttrxt660002mg2byiszzvr9", "published_at": "2026-09-09 07:28:23.107252+00:00", "updated_at": "2026-09-09 07:28:25.233414+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-agents", "ai-research"], "entities": ["arXiv"], "alternates": {"html": "https://wpnews.pro/news/merit-finds-memory-lifts-tool-agent-success-from-0-00-to-0-55-1-00", "markdown": "https://wpnews.pro/news/merit-finds-memory-lifts-tool-agent-success-from-0-00-to-0-55-1-00.md", "text": "https://wpnews.pro/news/merit-finds-memory-lifts-tool-agent-success-from-0-00-to-0-55-1-00.txt", "jsonld": "https://wpnews.pro/news/merit-finds-memory-lifts-tool-agent-success-from-0-00-to-0-55-1-00.jsonld"}}