Pith Number

pith:7KV2TX4S

pith:2026:7KV2TX4SGGWHL42GIHWMI5PL3P

not attested not anchored not stored refs resolved

Proxy Compression for Language Modeling

Lingpeng Kong, Lin Zheng, Qian Liu, Xiachong Feng, Xinyu Li

Proxy compression trains language models jointly on raw bytes and compressed sequences so they can use efficient inputs during training yet run purely on raw bytes at inference.

arxiv:2602.04289 v2 · 2026-02-04 · cs.CL · cs.LG

Open paper page JSON Open Graph Bundle Merged state Verified badge What is a Pith Number?

Add to your LaTeX paper

\usepackage{pith}
\pithnumber{7KV2TX4SGGWHL42GIHWMI5PL3P}

Prints a linked badge after your title and injects PDF metadata. Compiles on arXiv. Learn more · Embed verified badge

Record completeness

1 Bitcoin timestamp

2 Internet Archive

3 Author claim open · sign in to claim

4 Citations open

5 Replications open

✓ Portable graph bundle live · download bundle · merged state

The bundle contains the canonical record plus signed events. A mirror can host it anywhere and recompute the same current state with the deterministic merge algorithm.

Claims

C1strongest claim

proxy compression substantially improves training efficiency and significantly outperforms pure byte-level baselines given fixed compute budgets. As model scale increases, these gains become more pronounced, and proxy-trained models eventually match or surpass tokenizer approaches, all while operating solely on raw bytes.

C2weakest assumption

that the internal alignment learned during joint training on compressed and raw views transfers effectively to pure raw-byte inference without performance degradation or the need for continued compressed inputs.

C3one line summary

Proxy compression trains language models on both raw bytes and compressed sequences to enable efficient training with raw-byte inference at test time.

References

29 extracted · 29 resolved · 4 Pith anchors

[1] Magnet: Improving the multilingual fairness of language models with adaptive gradient-based tokenization 2023

[2] URL https://openreview.net/forum? id=PEpbUobfJv. Cao, K. and Rimell, L. You should evaluate your language model on marginal likelihood over tokeni- sations. InProceedings of the 2021 Conference on Emp 2021

[3] Evaluating Large Language Models Trained on Code 2021 · arXiv:2107.03374

[4] The Llama 3 Herd of Models 2019 · doi:10.18653/v1/n19-1423

[5] emnlp-industry.58/ 2023

Formal links

2 machine-checked theorem links

Cited by

5 papers in Pith

Efficient Pre-Training with Token Superposition

Decoupling the Benefits of Subword Tokenization for Language Model Training via Byte-level Simulation

Scratchpad Patching: Decoupling Compute from Patch Size in Byte-Level Language Models

Efficient Pre-Training with Token Superposition

Decoupling the Benefits of Subword Tokenization for Language Model Training via Byte-level Simulation

Receipt and verification

First computed	2026-05-17T23:39:00.067532Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

faaba9df9231ac75f34641ecc475ebdbe92a02fce573c3b2b23b5b3b73d1c5ce

Aliases

arxiv: 2602.04289 · arxiv_version: 2602.04289v2 · doi: 10.48550/arxiv.2602.04289 · pith_short_12: 7KV2TX4SGGWH · pith_short_16: 7KV2TX4SGGWHL42G · pith_short_8: 7KV2TX4S

Agent API

Resolver JSON Graph JSON Events JSON Schema Signing key

Verify this Pith Number yourself

curl -sH 'Accept: application/ld+json' https://pith.science/pith/7KV2TX4SGGWHL42GIHWMI5PL3P \
  | 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: faaba9df9231ac75f34641ecc475ebdbe92a02fce573c3b2b23b5b3b73d1c5ce

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "a4eb8756ec90079c47ff7deb3085703cf3dc7bc4f068cec585aad1704d97d8c9",
    "cross_cats_sorted": [
      "cs.LG"
    ],
    "license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
    "primary_cat": "cs.CL",
    "submitted_at": "2026-02-04T07:36:46Z",
    "title_canon_sha256": "e920092daf9b8b6cd485968f985685df05792529c32cedad61380bddef078911"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2602.04289",
    "kind": "arxiv",
    "version": 2
  }
}