Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language Models · HackerPapers discussions

Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language Models — mentioned by softwarewright

Thu, 03 Sep 2026 18:59:12 GMT

How does your Engram compare to the Engram in this paper? https://arxiv.org/abs/2601.07372

Paper and discussion history. A mention is not an endorsement.

Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language Models — mentioned by kzrdude

Thu, 27 Aug 2026 19:13:09 GMT

By the way; I found that DeepSeek says that it is even an "ideal complement to modern MoE architecture" https://arxiv.org/abs/2601.07372v1

Paper and discussion history. A mention is not an endorsement.

Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language Models — mentioned by arw0n

Fri, 27 Mar 2026 06:05:55 GMT

See also the new Deepseek paper on engram transformers for some progress in this area: https://arxiv.org/pdf/2601.07372v1 They observe significant gains in factual knowledge retrieval capabilities, but reasoning barely moves the needle.

Paper and discussion history. A mention is not an endorsement.

Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language Models — HN submission

Tue, 24 Mar 2026 20:16:33 GMT

DeepSeek: Conditional Memory via Scalable Lookup

7 points; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.

Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language Models — HN submission

Wed, 21 Jan 2026 05:08:33 GMT

Subject-based weight routing for LLMs (27 days before DeepSeek Engram)

I run LLM inference on an IBM POWER8 S824 with 576GB RAM – a $700 eBay server from 2014. In December 2025, I built "RAM Coffers" – banking model weights by subject domain with hot caching and resonance routing. On January 12, 2026, DeepSeek published "Engram" (arXiv:2601.07372) describing the same core idea: route queries to cached weight banks based on subject matter. The concepts are similar because I built it first. YouTube video from December 17, 2025: https://youtu.be/T_o39s7r0iE Terminal shows "RAM Coffers: ON | L2/L3 Resident: ON" – 26 days before their paper. Core shared concept…

1 point; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.

Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language Models — mentioned by AutoJanitor

Mon, 19 Jan 2026 19:15:06 GMT

I'm an independent researcher running LLMs on an IBM POWER8 server (320GB RAM, vintage 2015 hardware). On December 16, 2025, I developed "RAM Coffers" - a NUMA-distributed conditional memory system that selectively houses model weights in RAM banks with resonance-based routing for O(1) knowledge retrieval. On January 12, 2026, DeepSeek published their "Engram" paper (arXiv:2601.07372) describing the same core concept: separating static knowledge from dynamic computation via O(1) lookup. Same idea. 27 days apart. No connection. DOI: 10.6084/m9.figshare.31093429 GitHub…

Paper and discussion history. A mention is not an endorsement.

Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language Models — HN submission

Tue, 13 Jan 2026 20:55:40 GMT

Conditional Memory via Scalable Lookup

2 points; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.