Skip to content
OpenTrain AIFor AI Companies

HFEPX · Eval paper review

Physics of Language Models: Part 4.2, Canon Layers at Scale where Synthetic Pretraining Resonates in Reality

Zeyuan Allen-Zhu

Published

Nov 20, 2026

Citations

0

Trust level

Low

Usefulness score

5/100 (Low)

Extraction confidence

45% (Low)

Derived from extracted protocol signals and abstract evidence.

Rater population

Not reported

Signals refreshed

Nov 20, 2026

Should you rely on this paper?

This paper is adjacent to HFEPX scope and is best used for background context, not as a primary protocol reference.

Use this as background context only. Do not make protocol decisions from this page alone.

Best use

Background context only

Use if you need

A benchmark-and-metrics comparison anchor.

What to verify

Validate the evaluation procedure and quality controls in the full paper before operational use.

Main weakness

This paper looks adjacent to evaluation work, but not like a strong protocol reference.

Human feedback signal
Not explicit
Not explicit in abstract metadata
Evaluation signal
Detected
Eval setup described
Usefulness for eval research
5/100
Adjacent candidate

Treat as adjacent context, not a core eval-method reference.

Abstract

Do standard benchmarks see the architectural differences that matter? We study it with 64 from-scratch pretraining runs of Transformers and three linear-model families at 1.3-8B parameters on 1-2T Nemotron-CC tokens, at a cost of roughly 100k USD per 8B run. In academic-scale pretraining (1.3B parameters, 100B tokens), architectural differences can all but disappear when the models are carefully tuned. Instead, at 1-2T tokens they become measurable: Canon layers, which add local token mixing, consistently improve Transformers, saving about 30% of pretraining tokens for the same MMLU accuracy at 8B. We also introduce GLA5 and GDN2, which improve gated linear attention and Gated DeltaNet with low-rank output gates. Surprisingly, with Canon, the simplest linear model, gated linear attention, matches Gated DeltaNet and outperforms Mamba2 even at 8B/1T. Therefore, the extra machinery of Gated DeltaNet and Mamba2 may do little more than mix neighboring tokens. Even at 8B/1T, however, high scores on standard benchmarks can hide weaknesses in basic in-context skills. In particular, linear models that match or beat Transformers on these benchmarks cannot, in fact, reliably retrieve information from a short 500-token context. Moreover, no model, Transformer or linear, reliably beats random guessing on arguably the simplest two-hop question, stated in a few short sentences, even at this scale. These findings $\textbf{challenge traditional architecture research}$ based primarily on standard benchmark scores. In contrast, small synthetic playgrounds, which predicted the architectural differences above at <100M parameters, can reveal what these benchmarks miss. For practitioners, one lesson is to include in-context reasoning examples in pretraining data from an early stage, rather than reserve them for later mid-training, let alone post-training.

What we could verify

These are the protocol signals we could actually recover from the available paper metadata. Use them to decide whether this paper is worth deeper reading.

Human Feedback Types

missing

None explicit

No explicit feedback protocol extracted.

"Do standard benchmarks see the architectural differences that matter?"

Evaluation Modes

partial

Automatic Metrics

Includes extracted eval setup.

"Do standard benchmarks see the architectural differences that matter?"

Quality Controls

missing

Not reported

No explicit QC controls found.

"Do standard benchmarks see the architectural differences that matter?"

Benchmarks / Datasets

partial

MMLU

Useful for quick benchmark comparison.

"Instead, at 1-2T tokens they become measurable: Canon layers, which add local token mixing, consistently improve Transformers, saving about 30% of pretraining tokens for the same MMLU accuracy at 8B."

Reported Metrics

partial

Accuracy

Useful for evaluation criteria comparison.

"Instead, at 1-2T tokens they become measurable: Canon layers, which add local token mixing, consistently improve Transformers, saving about 30% of pretraining tokens for the same MMLU accuracy at 8B."

Benchmarks and datasets

MMLU

Reported metrics

accuracy
Human feedback details
Uses human feedback
No
Feedback types
None
Rater population
Not reported
Expertise required
General
Evaluation details
Evaluation modes
Automatic Metrics
Agentic eval
None
Quality controls
Not reported
Evidence quality
Low
Use this page as
Background context only

Research brief

Metadata summary

Do standard benchmarks see the architectural differences that matter?

Based on abstract + metadata only. Check the source paper before making high-confidence protocol decisions.

Key takeaways

  • Do standard benchmarks see the architectural differences that matter?
  • We study it with 64 from-scratch pretraining runs of Transformers and three linear-model families at 1.3-8B parameters on 1-2T Nemotron-CC tokens, at a cost of roughly 100k USD per 8B run.
  • In academic-scale pretraining (1.3B parameters, 100B tokens), architectural differences can all but disappear when the models are carefully tuned.

Researcher actions

  • Compare this paper against others mentioning MMLU.
  • Validate inferred eval signals (Automatic metrics) against the full paper.
  • Use related-paper links to find stronger protocol-specific references.

Caveats

  • Generated from abstract + metadata only; no PDF parsing.
  • Signals below are heuristic and may miss details reported outside the abstract.

Recommended queries

Contribution summary

  • Do standard benchmarks see the architectural differences that matter?
  • We study it with 64 from-scratch pretraining runs of Transformers and three linear-model families at 1.3-8B parameters on 1-2T Nemotron-CC tokens, at a cost of roughly 100k USD per 8B run.

Why it matters for eval

  • Even at 8B/1T, however, high scores on standard benchmarks can hide weaknesses in basic in-context skills.

Researcher checklist

  • Human feedback protocol is explicit

    No explicit human feedback protocol detected.

  • Evaluation mode is explicit

    Detected: Automatic Metrics

  • Quality control reporting appears

    No calibration/adjudication/IAA control explicitly detected.

  • Benchmark or dataset anchors are present

    Detected: MMLU

  • Metric reporting is present

    Detected: accuracy