AIINEWS

AIINEWS

Global intelligence on artificial intelligence and computing

Meta FAIR's self-optimized AI memory scores 56.0 across three tasks

AIINEWS Editorial Board•ai-machine-learning
20 November, 2026•2 min read
Meta FAIR's self-optimized AI memory scores 56.0 across three tasks

Executive Briefing

“Meta FAIR has introduced a method that lets an AI agent discover how to store and retrieve...”

Key Takeaways

  • 1.The method uses file-based memory. One instruction updates it after a subtask, while another retrieves notes before the next subtask.
  • 2.During the search stage, an agent rewrites the two memory instructions. Each candidate runs through the full task sequence five times, and selection uses a holdout set the agent cannot see.
  • 3.According to the article, the optimized instructions averaged 56.0 across Database, Codebase and Spectrum, versus 44.6 for the best human-written option. All reported result rows used Kimi-K2.6.
  • 4.The team also used runs with optimized instructions to fine-tune the model. When both versions used generic memory instructions, this distilled version averaged 43.1 across 10 matched seeds, versus 37.6 for fine-tuning on runs made with generic instructions.
  • 5.The search did not improve consistently with each revision. The article reports that 37 of 44 revisions failed to beat the previous best.

Meta FAIR has introduced a method that lets an AI agent discover how to store and retrieve information between subtasks. In the results reported by the article, the discovered instructions averaged 56.0 across three tasks, compared with 44.6 for the best human-written instructions.

Key Developments & Primary Facts

The method uses file-based memory. One instruction updates it after a subtask, while another retrieves notes before the next subtask.

During the search stage, an agent rewrites the two memory instructions. Each candidate runs through the full task sequence five times, and selection uses a holdout set the agent cannot see.

According to the article, the optimized instructions averaged 56.0 across Database, Codebase and Spectrum, versus 44.6 for the best human-written option. All reported result rows used Kimi-K2.6.

The team also used runs with optimized instructions to fine-tune the model. When both versions used generic memory instructions, this distilled version averaged 43.1 across 10 matched seeds, versus 37.6 for fine-tuning on runs made with generic instructions.

The search did not improve consistently with each revision. The article reports that 37 of 44 revisions failed to beat the previous best.

Strategic Analysis & Implications

The 56.0 score measures the optimized instructions, not the fine-tuned model. The 43.1 versus 37.6 result suggests fine-tuning retained some benefit after a return to generic instructions, but did not reproduce the optimized instructions' full score.

Because many revisions brought no improvement, testing and selection are central to the method, rather than simply having an agent write more instructions.

Outlook & Factors to Watch

The article does not disclose the training hardware, leaving the cost of reproducing the fine-tuning step unclear.

The tests use new instances of the same three tasks. They do not establish how the method would perform on different kinds of work.

Independently verified, edited, and published by AIINEWS Editorial Board
3-tier verified
Editorial standards
Share this intelligence report:
Expanded Perspectives & Intelligence

Related Dispatches & Deep Dives

Curated 8 specialized reports
Macmillan Staff Say the Publisher Threw a Party for a Workplace AI Assistant That Was Later Killed

Macmillan Staff Say the Publisher Threw a Party for a Workplace AI Assistant That Was Later Killed

As staff at the Lower Manhattan headquarters of Macmillan Publishers headed into work one morning in January 2024, they filed past something strange on the walls—flyers welcoming a new member of the team named “NOA.” The cafeteria was festooned with purple balloons and banners, according to several workers present, and with NOA’s name in light-up marquee letters. A plate full of doughnuts spelled out “NOA IS HERE.” NOA was an acronym for No Ordinary Assistant, which many of Macmillan’s human assistants found “offensive,” multiple employees tell WIRED. In an email to staff, NOA was introduced as an AI chatbot developed by the German software company Chaptr, which is “backed” by Holtzbrinck, Macmillan’s parent company.

8 min read7,354
Apple is hiring engineers to build its own AI agent
Artificial Intelligenceabout 4 hours ago

Apple is hiring engineers to build its own AI agent

Apple is recruiting software engineers to develop an autonomous AI assistant capable of browsing the web, working across applications, writing code, and calling services. This project indicates the technology company is moving toward expanding autonomous capabilities for Siri.

1 min read1,184

Explore AIINEWS Intelligence Vault

Over 2,000 in-depth dispatches structured across 18 specialized tech & AI categories.

Browse all 18 categories