Pith Number

pith:ZSYKFRMH

pith:2026:ZSYKFRMHFDPX3NF7JZYHDDB6RH

not attested not anchored not stored refs resolved

Runtime-Orchestrated Second-Order Optimization for Scalable LLM Training

Junhao Zhang, Wes Armour, Yishun Lu, Zeyu Yang

Asteria enables practical second-order LLM training by managing optimizer state and background tasks at the runtime level.

arxiv:2605.16184 v1 · 2026-05-15 · cs.DC · 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{ZSYKFRMHFDPX3NF7JZYHDDB6RH}

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

Our results suggest that second-order LLM training can be made practical not by simplifying the optimizer alone, but by rethinking how optimizer state, background computation, and distributed synchronization are managed at the runtime level.

C2weakest assumption

The bounded-staleness protocol and asynchronous shadow-state preparation preserve optimizer effectiveness without introducing unacceptable latency or convergence degradation; this is invoked in the description of the distributed training protocol and the training-hook mechanism.

C3one line summary

Asteria is a runtime system that enables second-order optimization for LLMs by dynamically distributing optimizer state across GPU, CPU, and NVMe while using asynchronous inverse-root computations and bounded-staleness synchronization.

References

30 extracted · 30 resolved · 13 Pith anchors

[1] Decoupled weight decay regularization

[2] Decoupled Weight Decay Regularization · doi:10.48550/arxiv.1711.05101

[3] Shampoo: Preconditioned stochastic tensor optimization 2018 · arXiv:1802.09568

[4] SOAP: Improving and Stabilizing Shampoo using Adam 2025 · doi:10.48550/arxiv.2409.11321

[5] Towards Learning Boulder Excavation with Hydraulic Excavators 2026 · doi:10.48550/arxiv.2

Formal links

2 machine-checked theorem links

Receipt and verification

First computed	2026-05-20T00:01:56.853619Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

ccb0a2c58728df7db4bf4e70718c3e89d3b36b62eaaedcecd3be058969adb794

Aliases

arxiv: 2605.16184 · arxiv_version: 2605.16184v1 · doi: 10.48550/arxiv.2605.16184 · pith_short_12: ZSYKFRMHFDPX · pith_short_16: ZSYKFRMHFDPX3NF7 · pith_short_8: ZSYKFRMH

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/ZSYKFRMHFDPX3NF7JZYHDDB6RH \
  | 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: ccb0a2c58728df7db4bf4e70718c3e89d3b36b62eaaedcecd3be058969adb794

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "2856b7d534b97f3037fc1f1c549071a6b52c9393ee1e0efd96d286bd79300681",
    "cross_cats_sorted": [
      "cs.LG"
    ],
    "license": "http://creativecommons.org/licenses/by/4.0/",
    "primary_cat": "cs.DC",
    "submitted_at": "2026-05-15T17:03:55Z",
    "title_canon_sha256": "10b51166bda78ff2dac4af4ebf02f7db72a61a7ba728a54249f0b5d3d643eb95"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2605.16184",
    "kind": "arxiv",
    "version": 1
  }
}