pith:6WJVDM3V
Rate-Distortion Optimization for Transformer Inference
A rate-distortion framework lets transformers compress intermediate representations to cut inference bitrate while preserving accuracy.
arxiv:2601.22002 v4 · 2026-01-29 · cs.LG · cs.IT · math.IT
Add to your LaTeX paper
\usepackage{pith}
\pithnumber{6WJVDM3VOOUXRUZLI5B223YRG3}
Prints a linked badge after your title and injects PDF metadata. Compiles on arXiv. Learn more · Embed verified badge
Record completeness
Claims
The simplest of the proposed codecs achieves substantial rate savings, outperforming more complex methods, and empirical rates for different architectures and tasks are driven by the derived information-theoretic bounds on achievable rate.
That the learned compression of intermediate representations preserves downstream task accuracy sufficiently for the rate-distortion tradeoff to remain useful in practice, without the paper specifying how accuracy degradation is measured across all possible splits.
A rate-distortion framework for lossy compression of transformer representations yields substantial bitrate savings on language tasks while preserving accuracy, with observed rates aligning to derived information-theoretic bounds.
Receipt and verification
| First computed | 2026-06-23T01:12:00.617065Z |
|---|---|
| Builder | pith-number-builder-2026-05-17-v1 |
| Signature | Pith Ed25519
(pith-v1-2026-05) · public key |
| Schema | pith-number/v1.0 |
Canonical hash
f59351b37573a978d32b4743ad6f1136fc8d2618d01cd9925b9b095cb2ad36eb
Aliases
· · · · ·Agent API
Verify this Pith Number yourself
curl -sH 'Accept: application/ld+json' https://pith.science/pith/6WJVDM3VOOUXRUZLI5B223YRG3 \
| jq -c '.canonical_record' \
| python3 -c "import sys,json,hashlib; b=json.dumps(json.loads(sys.stdin.read()), sort_keys=True, separators=(',',':'), ensure_ascii=False).encode(); print(hashlib.sha256(b).hexdigest())"
# expect: f59351b37573a978d32b4743ad6f1136fc8d2618d01cd9925b9b095cb2ad36eb
Canonical record JSON
{
"metadata": {
"abstract_canon_sha256": "cb79855f0642aa9df7a60d56a819c0024a5725f78487d7348e8bb0a9f3f15756",
"cross_cats_sorted": [
"cs.IT",
"math.IT"
],
"license": "http://creativecommons.org/licenses/by-sa/4.0/",
"primary_cat": "cs.LG",
"submitted_at": "2026-01-29T17:12:46Z",
"title_canon_sha256": "fa7ea1ba75428013a2c8ee3348e568df725787f7e69f5c14286ee8e88d834550"
},
"schema_version": "1.0",
"source": {
"id": "2601.22002",
"kind": "arxiv",
"version": 4
}
}