AI Preprint Pulse

Long-context inference optimizations (IndexCache + ... + BEAM + MemFactory + Omni-SimpleMem + AMA-Bench + ByteRover + OmniMEM + Neuro-Symbolic Dual Mem)

Long-context inference optimizations (IndexCache + ... + BEAM + MemFactory + Omni-SimpleMem + AMA-Bench + ByteRover + OmniMEM + Neuro-Symbolic Dual Mem)

Key Questions

What is Neuro-Symbolic Dual Memory?

Neuro-Symbolic Dual Memory decouples progress and feasibility memories to reduce agent drift in long-horizon tasks like ALFWorld and WebShop. It aligns progress tracking with feasibility assessment for better agent performance in extended contexts.

How does ByteRover improve long-horizon tasks?

ByteRover achieves 96.1% performance on long-horizon tasks through agent-native memory using LLM-curated hierarchical context. It enables efficient memory management for complex, extended agent interactions.

What is OmniMEM?

OmniMEM is a multimodal memory system designed for handling diverse data types in agent applications. It supports advanced memory augmentation as part of long-context optimizations.

What does AMA-Bench evaluate?

AMA-Bench is a benchmark for evaluating long-horizon memory in agentic applications. It provides standardized tests for memory performance in prolonged agent scenarios.

What is the BEAM benchmark?

BEAM is a memory benchmark involving 10M conversations, demonstrating that even 1M context windows are insufficient for robust agent memory. It highlights needs for advanced memory systems beyond long contexts.

Neuro-Symbolic Dual Mem decouples progress/feasibility mems to cut agent drift in ALFWorld/WebShop; ByteRover 96.1% long-horizon; OmniMEM multimodal; AMA-Bench evals; BEAM 10M convos; MemFactory GRPO 14.8%. Power/photonics/multi-agent benches urgent. Status: developing.

Sources (7)
Updated Apr 8, 2026
What is Neuro-Symbolic Dual Memory? - AI Preprint Pulse | NBot | nbot.ai