Textual Memory
Long-context, persona, script, and conversation-memory benchmarks.
THE JOURNAL
15 postsLatest stories
48 Hours Left: CSIG AML Cycle 2 Project Submission Deadline Approaches
Key dates for Full evaluation, resuming existing tasks, participant and team registration, result review, and the official Cycle 2 leaderboard release.
AML Is in Full Swing: Participant FAQ, Part 2
Practical answers on application follow-ups, version registration, cross-track integration, endpoint issues, and model eligibility for Cycle 2.
AML Phase II Is Open: Participant FAQ
AML Cycle 2 public evaluation opened on September 20, 2026, at 00:00. This FAQ covers registration, models, Add/Search integration, recovery, and results.
Two Days to Go | CSIG-Hosted Agent Memory Challenge 2026 Cycle 2 Participation Guide
Two Days to Go | Agent Memory Challenge 2026 Cycle 2 Participation Guide
AML Official Analysis II | Coding Memory Leaderboards: How Do Industry and Open-Source Methods Perform?
Compared with textual memory, Coding Memory asks a more direct question: can experience accumulated during previous development actually help an Agent complete subsequent software engineering tasks?
AML Leading-Method Spotlight | FlowGrid AML Retriever: How a Markdown Decision Log Became an Agent Memory Retriever
There is no single answer to Agent Memory.
AML Leading-Method Spotlight | ChronoHybridMem: From Conversations to Verifiable Memory Evidence
There is no single answer to Agent Memory.
AML Leading Methods Explained | ActiveMemoryIndex: Why Does a System Without Memory Governance Rank First in Governance?
ActiveMemoryIndex explains its raw-record-first design, transparent ablations, the limits of item-count budgets, and the next steps for neighboring context and temporal reasoning.
Hosted by CSIG | RMB 150,000 Prize Pool: The Second Agent Memory Challenge Is Coming
The second Agent Memory Challenge will open on September 20, 2026, with Textual, Coding, and Multimodal tracks, unified evaluation, and an RMB 150,000 prize pool.
AML Leading Methods Explained | ReFind: Active Search Lets Agents Browse Original Chat Logs
ReFind preserves raw chat logs and moves memory computation to query-time active search. Its AML results and the InMind benchmark reveal both the strengths and limits of this approach.
AML Leading Methods Explained | From Similarity to Completeness: How InvMem Retrieves Truly Useful Long-Term Memories
InvMem combines fine-grained chunking, Dense and BM25 retrieval, Weighted RRF, and same-session neighboring expansion to turn relevant fragments into complete memory evidence.
AML Official Analysis, Part I: What Does the Textual Open-Source Leaderboard Reveal?
The top three are separated by just 0.22 points, yet excel in different ways. A closer look at seven capability areas reveals distinctive strengths, weaknesses, and open questions beyond the overall ranking.
The First AML Leaderboard Is Officially Released!
The first AML leaderboard publishes textual-track results from 67 memory frameworks, evaluated and reviewed through a unified pipeline. Explore rankings, capability profiles, and opportunities to contribute.
An Agent Only Starts to Understand You When It Has a Yesterday
Agents need more than the ability to think: they need the ability to remember. AML establishes a shared standard for comparing memory systems and invites the community to the first Agent Memory Challenge.
The Agent Memory Challenge Launches: The First Open Tournament for Memory Systems
AML brings multi-source data, a unified evaluation pipeline, capability profiles, and low-barrier integration together in an open, reproducible evaluation framework. The first Agent Memory Challenge is now open.
No posts in this category yet
The First AML Leaderboard Is Officially Released!
Agent Memory Leaderboard
Today, the first Agent Memory Leaderboard (AML) is officially unveiled.
As agents move from single-turn question answering to long-term collaboration, memory is no longer merely an add-on. It is a foundational capability for continuously understanding users, accumulating experience, and completing complex tasks. Measuring this capability under consistent conditions has become a shared challenge for research and industry.
Jointly initiated by researchers and teams from more than 20 universities, research institutions, and open-source communities in China and abroad, AML is dedicated to building an open evaluation framework and a long-running public leaderboard for Agent Memory. The first evaluation covers both open-source methods and commercial products. Every system version admitted to the official leaderboard has completed testing and review under a unified pipeline and set of standards.
As of 12:00 on August 12, 2026, the official AML website had received more than 200,000 visits. The competition page alone exceeded 100,000 visits within ten days of launch. The challenge sparked lively discussion across online platforms and received registration applications from 136 teams.
🏆 The First Leaderboard Is Officially Published
Following rigorous screening and testing, 67 representative memory frameworks successfully completed the first evaluation. The initial leaderboard officially publishes results for the textual track, with separate participation channels for Open-source Methods and Commercial Products:
▍Open-source Methods Leaderboard [Textual Track]
▍Commercial Products Leaderboard [Textual Track]
For complete rankings, individual capability scores, and version information, open the leaderboard page through 'Read Original' at the end of the original article.
This leaderboard is officially published by AML. Rankings, scores, and conclusions in this article apply only within the corresponding participation division, system version, evaluation suite, and rules of the relevant cycle.
⚙️ One Unified Pipeline Makes Results Truly Comparable
In the past, different memory systems often used different datasets, answer models, and scoring methods, making public scores difficult to compare directly.
AML standardizes the core responsibilities of participating systems as Add (writing memories) and Search (retrieving memories). The platform centrally performs Answer (generating answers from retrieved results), Eval (scoring), result aggregation, and pre-publication review. This decoupling minimizes interference from different choices of answer models, scoring models, or statistical conventions, making evaluation results genuinely comparable and credible within clearly defined boundaries.
🧭 Look Beyond Rankings to Capability Structure
Memory capability is not simply information recall. For textual memory tasks, AML examines explicit fact recall, relational and multi-hop composition, temporal and event sequences, memory governance, personalization and care, rule and workflow execution, and epistemic safety and privacy.
Publishing the leaderboard is not the finish line, but a new starting point for exploring the foundations of memory. AML will continue with a series of in-depth analyses. We will examine representative technical architectures and thought-provoking experimental findings from this evaluation, reviewing scores across key capability areas. Our goal is to help developers look beyond the leaderboard's numbers and understand the underlying development and future direction of Agent Memory technology.
AML's official evaluation and leaderboard operate continuously, and evaluation remains open. We invite academic research teams and industry developers worldwide to submit systems. We also warmly welcome new benchmark datasets and evaluation tasks from the community, so that together we can broaden the standards for assessing agent memory capabilities!
We will contact teams selected for the first AML challenge's leaderboard rewards and Community Contribution Program rewards by email before August 31. Please watch your inbox.
Thank you to all participating teams, contributors of data and evaluation methods, and researchers and developers following the progress of Agent Memory!
Make memory capabilities measurable. Make technical progress verifiable.
📧 Contact and collaboration: you are welcome to reach the evaluation team through GitHub Discussion or our official email address.