Reference library
LLM-Pruner: On the Structural Pruning of Large Language Models
Paper summary
LLM-Pruner removes coupled structural components using gradient-based importance and then recovers capability with limited additional training. It targets practical model compression while minimizing dependence on the original training corpus.
Relevance to the project
A reference for structural rather than individual-weight pruning. May inform later cost-aware removal baselines, but changing simulation graphs require independent functional tests.
Compression with recovery training differs from lifetime local rewiring.
Source record
- Reference ID
- llm-pruner
- Archived version
- 2305.11627v3
- Summary provenance
- reference-library/papers/llm-pruner/metadata.json · summary
- Review status
- initial orientation summary; not a critical full-paper review
- Reading scope
- Selected abstract/opening and available closing discussion/conclusion passages in the preserved text extract; broader project notes provide context. No reproduction of experiments, exhaustive methods review, or equation/figure audit.