AITOPAI热报
PapersSelected73°

LatentPress context compression method

LatentPress: Context Compression Beyond Text and Vision

arXiv cs.LG@Zhengze Zhou, Hejian Sang
Paper· 2 sources
Read the original
Why we picked it

LatentPress processes context 5-9× faster than raw methods, offering significant efficiency gains for developers working with large language models.

LatentPress compresses conversational histories and long documents into continuous memory tokens that a frozen decoder reads directly. The method achieves 7.70× compression with 0.504 accuracy on LongMemEval, outperforming text summaries (0.184) and OCR-based compression (0.426 to 0.312). Writing takes 43ms per conversation, roughly 10× faster than text summarization or OCR reconstruction.

Original · arXiv cs.LG

LatentPress: Context Compression Beyond Text and Vision

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: https://github.com/xuyd16ai/context_softtoken_compress .

#LatentPress#LongMemEval#LongBench-QA#context compression#memory tokens

Keep reading

Recent stories with the same tag or section
PapersBuildersSelected20:32
JD ROAD MOPD refines multi-teacher distillation to cut gradient conflict

The frozen reference model makes agreement checks brittle: if it scores a token lower than both teachers, the teachers count as "agreeing" even when their gap is large, and in-between scores flip the logic. That limitation matters for anyone building multi-teacher distillation pipelines and evaluating whether relative-probability gating or explicit KL divergence between teachers gives a cleaner compatibility signal. Chinese lab output on distillation tricks like this tends to preview what lands in later open-source model training recipes.

#JD#MOPD#distillation#on-policy distillation
Teortaxes@teortaxesTexSource
PapersBuildersSelected17:16
Tencent Hunyuan releases WorkForge for training work agents

The dual programmatic-plus-semantic verifier design addresses the hardest part of agent training: rewarding long-horizon tasks without human grading. Teams building enterprise work agents can reuse the recipe to generate their own verifiable task suites from internal workflows rather than hand-crafted benchmarks. It sits alongside agent-environment efforts like task synthesis pipelines from other labs but anchors everything in expert-derived, evidence-backed tasks.

#Tencent Hunyuan#WorkForge#agents#open-source models
官方账号腾讯混元@TencentHunyuanSource