AI Coding Live News

Conversational memory for LLM agents traditionally divides into two opposing camps: expensive pipeline extraction that distills dialogues into structured factual memory stores versus raw conversational turn retrieval. Literature has reported contradictory findings. Researchers conducted a pre-registered, double-blind controlled study on held-out LoCoMo and LongMemEval benchmarks (arXiv:2609.34227). The investigation reveals that under constrained context budgets, selecting raw conversational turns via a lightweight typed decision model (Jev) is statistically non-inferior to full LLM fact extraction pipelines (95% one-sided bound -3.0 points vs. -5 margin)—while reducing write-time computational costs by a massive 3,061 times. Furthermore, the authors resolve literature contradictions by proving that reranking gains scale inversely with token budgets, while aggressive reranking degrades an agent's correct abstention reliability. All study protocols, data, and code are open-sourced at github.com/ris3abh/Engram.

LatentPress stores conversational histories and long documents as continuous memory tokens a frozen decoder reads through its input-embedding interface—no text reconstruction at inference. A small reader-matched writer (4.2M–26.2M adapter params, ~0.1% of the decoder) compresses 4–16×. On LongMemEval it reaches 0.504 accuracy at 7.70× compression vs 0.490 uncompressed, beating text summaries and OCR; writing is ~43ms per conversation and reading is 5–9× faster than raw context—useful for long-context agent memory.