Skip to content
OpenTrain AIFor AI Companies

HFEPX · Eval paper review

Fragile Knowledge, Robust Instruction-Following: The Width Pruning Dichotomy in Llama-3.2

Pere Martra

Published

Dec 27, 2025

Citations

0

Trust level

Moderate

Usefulness score

25/100 (Low)

Extraction confidence

55% (Moderate)

Derived from extracted protocol signals and abstract evidence.

Rater population

Not reported

Signals refreshed

May 6, 2026

Open arXiv Find implementation Shortlist (0) RSS feed

Tags

Automatic Metrics Long Horizon Math

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 for comparison and orientation, not as your only source.

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

No major weakness surfaced.

Human feedback signal
Not explicit
Not explicit in abstract metadata
Evaluation signal
Detected
Eval setup described
Usefulness for eval research
25/100
Adjacent candidate

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

If you are doing eval pipeline work, start here

Human Eval Papers LLM-as-Judge Papers Pairwise Preference Papers Tool-Use Evaluation Search

Abstract

Structured width pruning of GLU-MLP layers, guided by the Maximum Absolute Weight (MAW) criterion, reveals a systematic dichotomy in how reducing the expansion ratio affects different model capabilities. While performance on tasks relying on parametric knowledge (e.g., MMLU, GSM8K) and perplexity metrics degrades predictably, instruction-following capabilities improve substantially (+46% to +75% in IFEval for Llama-3.2-1B and 3B models), and multi-step reasoning remains robust (MUSR). This pattern challenges the prevailing assumption that pruning induces uniform degradation. We evaluated seven expansion ratio configurations using comprehensive benchmarks assessing factual knowledge, mathematical reasoning, language comprehension, instruction-following, and truthfulness. Our analysis identifies the expansion ratio as a critical architectural parameter that selectively modulates cognitive capabilities, rather than merely serving as a compression metric. We provide the first systematic characterization of this selective preservation phenomenon. Notably, we document a robust inverse correlation (r = -0.864, p = 0.012 in Llama-3B) between factual knowledge capacity (MMLU) and truthfulness metrics (TruthfulQA-MC2): as knowledge degrades, the model's ability to discriminate misconceptions improves consistently. This connects two previously distinct research areas, demonstrating that MAW-guided width pruning acts as a selective filter, reducing parametric knowledge while preserving or enhancing behavioral alignment. Additionally, we quantify context-dependent efficiency trade-offs: pruned configurations achieve up to 23% reduction in energy consumption (J/token) but incur penalties in single-request latency, whereas batch processing workloads benefit uniformly.

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.

"Structured width pruning of GLU-MLP layers, guided by the Maximum Absolute Weight (MAW) criterion, reveals a systematic dichotomy in how reducing the expansion ratio affects different model capabilities."

Evaluation Modes

strong

Automatic Metrics

Includes extracted eval setup.

"Structured width pruning of GLU-MLP layers, guided by the Maximum Absolute Weight (MAW) criterion, reveals a systematic dichotomy in how reducing the expansion ratio affects different model capabilities."

Quality Controls

missing

Not reported

No explicit QC controls found.

"Structured width pruning of GLU-MLP layers, guided by the Maximum Absolute Weight (MAW) criterion, reveals a systematic dichotomy in how reducing the expansion ratio affects different model capabilities."

Benchmarks / Datasets

strong

MMLU, GSM8K, IFEval, TruthfulQA

Useful for quick benchmark comparison.

"While performance on tasks relying on parametric knowledge (e.g., MMLU, GSM8K) and perplexity metrics degrades predictably, instruction-following capabilities improve substantially (+46% to +75% in IFEval for Llama-3.2-1B and 3B models), and multi-step reasoning remains robust (MUSR)."

Reported Metrics

strong

Perplexity

Useful for evaluation criteria comparison.

"While performance on tasks relying on parametric knowledge (e.g., MMLU, GSM8K) and perplexity metrics degrades predictably, instruction-following capabilities improve substantially (+46% to +75% in IFEval for Llama-3.2-1B and 3B models), and multi-step reasoning remains robust (MUSR)."

Benchmarks and datasets

MMLUGSM8KIFEvalTruthfulQA

Reported metrics

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

Research brief

Metadata summary

Structured width pruning of GLU-MLP layers, guided by the Maximum Absolute Weight (MAW) criterion, reveals a systematic dichotomy in how reducing the expansion ratio affects different model capabilities.

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

Key takeaways

  • Structured width pruning of GLU-MLP layers, guided by the Maximum Absolute Weight (MAW) criterion, reveals a systematic dichotomy in how reducing the expansion ratio affects different model capabilities.
  • While performance on tasks relying on parametric knowledge (e.g., MMLU, GSM8K) and perplexity metrics degrades predictably, instruction-following capabilities improve substantially (+46% to +75% in IFEval for Llama-3.2-1B and 3B models), and multi-step reasoning remains robust (MUSR).
  • This pattern challenges the prevailing assumption that pruning induces uniform degradation.

Researcher actions

  • Compare this paper against others mentioning MMLU and GSM8K.
  • Validate inferred eval signals (Automatic metrics, Long-horizon tasks) 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

MMLU GSM8K

Contribution summary

  • While performance on tasks relying on parametric knowledge (e.g., MMLU, GSM8K) and perplexity metrics degrades predictably, instruction-following capabilities improve substantially (+46% to +75% in IFEval for Llama-3.2-1B and 3B models),…
  • We evaluated seven expansion ratio configurations using comprehensive benchmarks assessing factual knowledge, mathematical reasoning, language comprehension, instruction-following, and truthfulness.
  • Additionally, we quantify context-dependent efficiency trade-offs: pruned configurations achieve up to 23% reduction in energy consumption (J/token) but incur penalties in single-request latency, whereas batch processing workloads benefit…

Why it matters for eval

  • We evaluated seven expansion ratio configurations using comprehensive benchmarks assessing factual knowledge, mathematical reasoning, language comprehension, instruction-following, and truthfulness.

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, GSM8K, IFEval, TruthfulQA

  • Metric reporting is present

    Detected: perplexity