analiza aiqbench
S3Mem introduces a query-time memory interface that structures heterogeneous histories into scene-event units to route compact evidence packs without requiring reader fine-tuning. The approach achieves a strong score-token trade-off, significantly reducing the number of evidence tokens needed for long-horizon question answering while maintaining competitive accuracy on benchmarks like LoCoMo and EMemBench. This method is particularly effective for tasks requiring localized event, state, temporal, and causal reasoning, offering a more efficient alternative to existing context expansion or graph neighborhood retrieval strategies.
S3Mem: Structured Spatiotemporal Scene-Event Memory for Long-Horizon Interactive Question Answering
arXiv:2605.28831v2 Announce Type: replace Abstract: Long-horizon memory question answering often requires sparse evidence from heterogeneous histories, including events, object states, visual observations, temporal relations, and causal steps. Existing memory interfaces expand reader context, retrieve semantically related chunks, or expose graph neighborhoods, but they are not explicitly designed to select compact evidence for a fixed reader. We propose Structured Spatiotemporal Scene--Event Memory (S3Mem), a query-time memory interface that writes textual, visual, and agent-use histories into structured scene--event units and routes compact evidence packs to the reader. Its router scores candidate units, query anchors, and anchor--support links, enabling both single-hop selection and short multi-hop evidence chains without reader fine-tuning or test-time training. Across LoCoMo, EMemBench Visual Games, and AMA-Bench, S3Mem provides a strong score--token trade-off, with the clearest gains on localized event, state, temporal, causal, or provenance evidence. On LoCoMo, S3Mem reaches \(0.48\) F1 and \(0.40\) BLEU with (1{,}073) evidence tokens per question, about \(15.8\times\) fewer than the LoCoMo reference. On EMemBench Visual Games, it obtains the best F1 and second-best accuracy with only \(189\)tokens.On AMA-Bench, it is not the highest-scoring method, but remains competitive while using the fewest reader-visible evidence tokens.
Abstract:Long-horizon memory question answering often requires sparse evidence from heterogeneous histories, including events, object states, visual observations, temporal relations, and causal steps. Existing memory interfaces expand reader context, retrieve semantically related chunks, or expose graph neighborhoods, but they are not explicitly designed to select compact evidence for a fixed reader. We propose Structured Spatiotemporal Scene--Event Memory (S3Mem), a query-time memory interface that writes textual, visual, and agent-use histories into structured scene--event units and routes compact evidence packs to the reader. Its router scores candidate units, query anchors, and anchor--support links, enabling both single-hop selection and short multi-hop evidence chains without reader fine-tuning or test-time training. Across LoCoMo, EMemBench Visual Games, and AMA-Bench, S3Mem provides a strong score--token trade-off, with the clearest gains on localized event, state, temporal, causal, or provenance evidence. On LoCoMo, S3Mem reaches \(0.48\) F1 and \(0.40\) BLEU with (1{,}073) evidence tokens per question, about \(15.8\times\) fewer than the LoCoMo reference. On EMemBench Visual Games, it obtains the best F1 and second-best accuracy with only \(189\)this http URL AMA-Bench, it is not the highest-scoring method, but remains competitive while using the fewest reader-visible evidence tokens.
Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
arXiv:2605.28831 [cs.CL]
(or arXiv:2605.28831v2 [cs.CL] for this version)
Text integral sub licența deschisă a publicației · 202 cuvinte
Subiecte conexe
[AINews] Claude Fable/Mythos 5.1: new SOTA model, 75% cache price cut but 70% more output tokens
Latent Space · 2 sept. 2026
[AINews] Fal’s H3 Max Live breaks the infinite videogen barrier
Latent Space · 1 sept. 2026
[AINews] The Future of Latent Space
Latent Space · 25 sept. 2026
AtomWorld-Mem: Memory-Restored World States for Long-Horizon Atomistic Evolution
arXiv · 28 sept. 2026