The agent memory problem just got a shared test. the agent memory challenge, opening its second…
the agent memory problem just got a shared test. the agent memory challenge, opening its second cycle september 20, grades whether agents notice when an old fact stops being true, after a first round drew 136 teams.
stale memory is worse than no memory.
Context
The Agent Memory Leaderboard's challenge page says the second Agent Memory Challenge is hosted by the China Society of Image and Graphics on its platform, with cycle 2 opening at 00:00 UTC+8 on September 20, a prize pool of 150,000 yuan for an open-source method leaderboard and a submission deadline of 2026.10.31. Participants provide standardized Add and Search APIs and the platform runs answering, evaluation and the leaderboard, with tracks including a Textual Memory track covering temporal events. A press release of September 28, 2026 from the leaderboard says cycle 1 drew more than 100 teams.
The 136 teams figure was not found in these texts, but more than 100 teams is compatible with it as a lower bound and does not refute it. That the challenge grades whether agents notice when an old fact stops being true is not stated in the text read, and the temporal events track is the nearest listed item, so the mechanism is unverified. This is a common benchmark and leaderboard and not shared agent memory. The challenge page is in Chinese and the English release describes a global opening on September 28. Stale memory is worse than no memory is the author's take.
Watch next
- The cycle 2 leaderboard results.
Sources
- Agent Memory Leaderboard: competitionagentmemoryleaderboard.ai
- GlobeNewswire: Agent Memory Challenge cycle 2 opens globally (September 28, 2026)globenewswire.com
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 22 September 2026 at 23:02 IST. Sources are the papers and datasets the note draws on.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →