Search papers, labs, and topics across Lattice.
The paper introduces WorldMemArena, a benchmark for evaluating multimodal agent memory in long-horizon tasks, focusing on the Action-World Interaction Loop with a four-stage lifecycle. It enables detailed diagnosis of memory failures across writing, maintenance, retrieval, and usage, and facilitates comparison of different memory architectures like RAG and harness-based systems. Experiments reveal that improved memory writing doesn't guarantee better performance, multimodal memory struggles with visual evidence, systems are unstable across domains, and harness memory, while flexible, is costly and unreliable.
Just because your agent can write and store memories well doesn't mean it can actually *use* them effectively in a dynamic, multimodal world.
Multimodal large language models are increasingly deployed as long-horizon agents, where memory must do more than recall: it must track an evolving world, revise what has gone stale, and surface the right evidence at decision time. Existing benchmarks measure recall over static dialogue, collapse memory into a single end-of-task accuracy, and reduce visual observations to captions, leaving us unable to localize failures to writing, maintenance, retrieval, or use. The rise of agent harnesses that author their own memory sharpens this gap, since we have no principled way to compare hand-designed pipelines with self-managing alternatives. To close these gaps, we formulate multimodal agent memory as an Action-World Interaction Loop with an observable four-stage lifecycle, and instantiate it in WorldMemArena: 400 multi-session multimodal tasks spanning Lifelong Evolution (evolving personal and task states) and Agentic Execution (memory from real observations, actions, and feedback), annotated with gold memory points, updates, distractors, and evidence chains for stage-level diagnosis. This enables the first head-to-head comparison of long-context, manually designed (RAG and external memory systems), and harness-based memory agents. Results show that: (1) better memory writing and storage do not guarantee better performance; (2) multimodal memory still struggles to fully use visual evidence; (3) systems are unstable across domains and degrade on realistic agentic trajectories; and (4) harness memory is more flexible but remains costly and less reliable.
