---
type: "raw_signal"
id: "built-on-gsq-and-rco-methods-from-ista-01406b"
slug: "built-on-gsq-and-rco-methods-from-ista-01406b"
title: "Built on GSQ and RCO methods from ISTA-DASLab with Swift-specific refinement"
description: "## GSQ proposes learned scalar quantization for low-bit LLM deployment"
retrieval_nugget: "## GSQ proposes learned scalar quantization for low-bit LLM deployment"
observed_at: "2026-10-12"
record_date: "2026-10-12"
date_kind: "observed_at"
why_it_matters: "Source: https://arxiv.org/abs/2604.18556"
novelty: "notable"
verification_level: "source-linked"
signal_type: "research"
source_platform: "arxiv.org"
topics: []
entities: ["GSQ","Llama 3.1","QTIP","Kimi-K2.5","ISTA-DASLab"]
related_patterns: []
source_url: "https://arxiv.org/abs/2604.18556"
source_urls: ["https://arxiv.org/abs/2604.18556"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "signal:built-on-gsq-and-rco-methods-from-ista-01406b"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/signals/built-on-gsq-and-rco-methods-from-ista-01406b/","markdown":"https://newruntime.com/signals/built-on-gsq-and-rco-methods-from-ista-01406b.md","json":"https://newruntime.com/signals/built-on-gsq-and-rco-methods-from-ista-01406b.json"}
---

# Built on GSQ and RCO methods from ISTA-DASLab with Swift-specific refinement

## Retrieval answer

## GSQ proposes learned scalar quantization for low-bit LLM deployment

## GSQ proposes learned scalar quantization for low-bit LLM deployment

An arXiv paper introduces Gumbel-Softmax Quantization (GSQ), a post-training scalar quantization method that jointly learns per-coordinate grid assignments and per-group scales. The authors report that it closes most of the accuracy gap to QTIP at 2 and 3 bits on Llama 3.1 8B and 70B Instruct, while using a symmetric scalar grid compatible with existing scalar inference kernels.

The paper also reports applying GSQ to public GGUF checkpoints and projecting results back into the same deployment format, as well as scaling the method to Kimi-K2.5, a trillion-scale mixture-of-experts model. These are reported experiments, not independent replication. The supplied source does not establish end-to-end latency, memory use, or quality in particular application workloads.

Source: https://arxiv.org/abs/2604.18556

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- collection: [Raw Signals](https://newruntime.com/signals.json): Open every source-linked raw signal.
- collection: [Source Ledger](https://newruntime.com/source-ledger.json): Inspect the public source evidence graph.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)