Pith Number

pith:P42XLZOB

pith:2025:P42XLZOB6RLJ2G7GSHYN375FG2

not attested not anchored not stored refs pending

EAGLE-3: Scaling up Inference Acceleration of Large Language Models via Training-Time Test

Chao Zhang, Fangyun Wei, Hongyang Zhang, Yuhui Li

By switching to direct token prediction and multi-layer feature fusion, EAGLE-3 enables draft models to improve with increased training data for faster LLM inference.

arxiv:2503.01840 v3 · 2025-03-03 · cs.CL

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

Add to your LaTeX paper

\usepackage{pith}
\pithnumber{P42XLZOB6RLJ2G7GSHYN375FG2}

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

EAGLE-3 abandons feature prediction in favor of direct token prediction and replaces reliance on top-layer features with multi-layer feature fusion via training-time test, significantly enhancing performance and enabling the draft model to fully benefit from scaling up training data, achieving up to 6.5x speedup and 1.4x over EAGLE-2.

C2weakest assumption

That direct token prediction combined with multi-layer feature fusion via training-time test will remove the constraints that previously limited gains from scaling training data, without introducing new accuracy or stability issues in the draft model.

C3one line summary

EAGLE-3 reaches up to 6.5x LLM inference speedup by replacing feature prediction with direct token prediction and multi-layer fusion through training-time test.

Formal links

1 machine-checked theorem link

Cited by

34 papers in Pith

SpecBlock: Block-Iterative Speculative Decoding with Dynamic Tree Drafting

Enhancing Cloud Network Resilience via a Robust LLM-Empowered Multi-Agent Reinforcement Learning Framework

ETS: Energy-Guided Test-Time Scaling for Training-Free RL Alignment

SSV: Sparse Speculative Verification for Efficient LLM Inference

Test-Time Speculation

Receipt and verification

First computed	2026-05-17T23:38:50.539247Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

7f3575e5c1f4569d1be691f0ddffa53688acad4a13f73e1dd00d7f19ef719af9

Aliases

arxiv: 2503.01840 · arxiv_version: 2503.01840v3 · doi: 10.48550/arxiv.2503.01840 · pith_short_12: P42XLZOB6RLJ · pith_short_16: P42XLZOB6RLJ2G7G · pith_short_8: P42XLZOB

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/P42XLZOB6RLJ2G7GSHYN375FG2 \
  | 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: 7f3575e5c1f4569d1be691f0ddffa53688acad4a13f73e1dd00d7f19ef719af9

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "d96088f2e47dac06a0977b5581508b768372ecfe4e76df8f9b2ab36b7028e971",
    "cross_cats_sorted": [],
    "license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
    "primary_cat": "cs.CL",
    "submitted_at": "2025-03-03T18:59:04Z",
    "title_canon_sha256": "ed257265660a32c954a9acae0b22dea0f1f57b9ae342a852e7d66a928032e253"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2503.01840",
    "kind": "arxiv",
    "version": 3
  }
}