LatentPress：超越文本与视觉的上下文压缩

View PDF HTML (experimental)

Abstract:Compressed context is usually carried as human-readable text or as rendered images that must be decoded, even when its consumer is a language model. We introduce LatentPress, which writes conversational histories and long documents into a third representation: continuous memory tokens that a frozen decoder reads directly through its input-embedding interface, with no text reconstruction at inference. A small reader-matched writer compresses $4$-$16\times$ while training only an adapter (4.2M-26.2M parameters, $\sim\!0.1\%$ of the decoder). On LongMemEval, LatentPress reaches $0.504$ accuracy at $7.70\times$ compression versus $0.490$ for uncompressed evidence, outperforming text summaries (0.184) and OCR-based compression (0.426 to 0.312). On LongBench-QA, in-domain writers match or exceed raw-context reading at $4$-$8\times$ compression, while $16\times$ trails raw. Writing takes 43ms per conversation, roughly an order of magnitude faster than text summarization or OCR reconstruction, and reading is $5$-$9\times$ faster than raw context or cached OCR. We validate the interface under two transfer settings, zero-shot from UltraChat to LongMemEval memory QA and from LongMemEval-derived QA to unseen LongBench document domains, establishing direct soft tokens as a practical machine-facing context interface beyond text and vision. The implementation of the experiments could be found at: this https URL.

Subjects:	Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
Cite as:	arXiv:2609.01507 [cs.LG]
	(or arXiv:2609.01507v3 [cs.LG] for this version)
	https://doi.org/10.48550/arXiv.2609.01507 arXiv-issued DOI via DataCite

Submission history

From: Hejian Sang [view email]
[v1] Tue, 1 Sep 2026 16:38:15 UTC (222 KB)
[v2] Thu, 3 Sep 2026 03:34:01 UTC (222 KB)
[v3] Sat, 10 Oct 2026 04:40:23 UTC (221 KB)