EverOS Hits SOTA Performance on LoCoMo
EverOS Hits SOTA Performance on LoCoMo
EverOS is an intelligent memory operating system designed to give AI the ability not just to remember, but to understand, reason, and evolve. On the LoCoMo benchmark, our approach built upon EverOS achieved a 92.3% reasoning accuracy (evaluated by LLM-Judge), outperforming comparable methods in our internal evaluation.
EverMind researchers
About 3 minutes to read

Most existing AI systems suffer from contextual forgetfulness — they can handle one conversation perfectly but fail to connect it with what came before. As AI becomes a long-term collaborator instead of a short-term assistant, such amnesia limits its capacity for real understanding and continuity. Our motivation in creating EverOS was simple yet ambitious: To transform AI memory from a passive data store into an active cognitive layer — one that captures meaning, maintains coherence, and builds a long-term understanding of users and tasks. We believe true intelligence requires the ability to remember with purpose and act with insight.
At its core, EverOS integrates layered memory extraction, structured knowledge organization, and adaptive retrieval mechanisms, enabling AI to build narratives from scattered interactions and make reasoning decisions grounded in accumulated understanding. These design principles come together in what we call our Unique Advantages — the foundation that makes EverOS not just a system of memory, but a system of foresight. Our unique advantages lie in three key dimensions that redefine how AI perceives, connects, and grows through memory.
Coherent Narrative
Beyond fragments, connecting stories. EverOS automatically links conversation pieces to form clear thematic contexts, enabling AI to truly understand. It naturally distinguishes between threads such as “Project A progress discussions” and “Team B strategy planning”, maintaining coherent logic within each theme. From scattered phrases to complete narratives, AI no longer just “understands one sentence” — it “understands the whole story.”
Evidence-Based Perception
Beyond retrieval, towards intelligent perception. EverOS proactively captures deep relationships between memories and tasks, allowing AI to “think thoroughly” at critical moments. For instance, when a user asks for food recommendations, the system recalls that “you had dental surgery two days ago” and adapts its suggestions accordingly. This is Contextual Awareness — intelligence built on understanding rather than isolated responses.
Living Profiles
Beyond records, towards dynamic growth. EverOS continuously updates user profiles in real time, learning from every interaction to know users more authentically. Over time, preferences, tone, and focus areas evolve naturally. The system doesn’t just “remember what you said,” it’s actively “learning who you are.”
Based on the EverOS framework and the GPT-4.1-mini model, we evaluated long-context memory reasoning performance on the LoCoMo dataset. According to assessments by LLM-Judge, our approach achieved a reasoning accuracy of 92.3%, outperforming comparable methods. The detailed experimental results are shown in the table below.

Most existing AI systems suffer from contextual forgetfulness — they can handle one conversation perfectly but fail to connect it with what came before. As AI becomes a long-term collaborator instead of a short-term assistant, such amnesia limits its capacity for real understanding and continuity. Our motivation in creating EverOS was simple yet ambitious: To transform AI memory from a passive data store into an active cognitive layer — one that captures meaning, maintains coherence, and builds a long-term understanding of users and tasks. We believe true intelligence requires the ability to remember with purpose and act with insight.
At its core, EverOS integrates layered memory extraction, structured knowledge organization, and adaptive retrieval mechanisms, enabling AI to build narratives from scattered interactions and make reasoning decisions grounded in accumulated understanding. These design principles come together in what we call our Unique Advantages — the foundation that makes EverOS not just a system of memory, but a system of foresight. Our unique advantages lie in three key dimensions that redefine how AI perceives, connects, and grows through memory.
Coherent Narrative
Beyond fragments, connecting stories. EverOS automatically links conversation pieces to form clear thematic contexts, enabling AI to truly understand. It naturally distinguishes between threads such as “Project A progress discussions” and “Team B strategy planning”, maintaining coherent logic within each theme. From scattered phrases to complete narratives, AI no longer just “understands one sentence” — it “understands the whole story.”
Evidence-Based Perception
Beyond retrieval, towards intelligent perception. EverOS proactively captures deep relationships between memories and tasks, allowing AI to “think thoroughly” at critical moments. For instance, when a user asks for food recommendations, the system recalls that “you had dental surgery two days ago” and adapts its suggestions accordingly. This is Contextual Awareness — intelligence built on understanding rather than isolated responses.
Living Profiles
Beyond records, towards dynamic growth. EverOS continuously updates user profiles in real time, learning from every interaction to know users more authentically. Over time, preferences, tone, and focus areas evolve naturally. The system doesn’t just “remember what you said,” it’s actively “learning who you are.”
Based on the EverOS framework and the GPT-4.1-mini model, we evaluated long-context memory reasoning performance on the LoCoMo dataset. According to assessments by LLM-Judge, our approach achieved a reasoning accuracy of 92.3%, outperforming comparable methods. The detailed experimental results are shown in the table below.

You may also like these
Related

Do public SKILL.md files actually make agents better?
SkillCorpus,SKILL.md,agent skills,skill curation,skill retrieval,LLM agents,SkillsBench,GDPVal,agent harness

CRAFT: learning how to fuse video tokens, not just which to drop
CRAFT,video token compression,vision-language models,video VLM,KV cache,prefill cost,token merging,token pruning,temporal reasoning

Self-evolving agents have a measurement problem
self-evolving agents,agent harness,HarnessBank,credit assignment,LLM agents,agent evaluation,harness optimization,significance testing

Skill Hub: a measured foundation for community-powered agents
skillhub,skill benchmark,SKILL.md,community skills,ai agent
EverOS Hits SOTA Performance on LoCoMo
EverOS is an intelligent memory operating system designed to give AI the ability not just to remember, but to understand, reason, and evolve. On the LoCoMo benchmark, our approach built upon EverOS achieved a 92.3% reasoning accuracy (evaluated by LLM-Judge), outperforming comparable methods in our internal evaluation.
EverMind researchers
About 3 minutes to read
EverMind
A straightforward solution to long-term coherence
Scan to join the community

Discord

© 2026 EverMind Team.
EverMind
A straightforward solution to long-term coherence
Scan to join the community

Discord

© 2026 EverMind Team.
EverMind
A straightforward solution to long-term coherence
Scan to join the community

Discord

© 2026 EverMind Team.