Reference library
TIES-Merging: Resolving Interference When Merging Models
Paper summary
TIES-Merging combines fine-tuned models by trimming small parameter changes, resolving sign conflicts, and aggregating aligned updates. Experiments show reduced interference compared with simpler model-merging approaches.
Relevance to the project
A comparison for combining externally learned mechanism candidates or optional LLM adaptations. Parameter correspondence and shared origin are essential before treating merging as recombination.
Model merging is not automatically useful for evolved builders or recurrent dynamics.
Source record
- Reference ID
- ties-merging
- Archived version
- 2306.01708v2
- Summary provenance
- reference-library/papers/ties-merging/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.