REVIEW 12 cited by
Model merging with SVD to tie the Knots
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Model merging with SVD to tie the Knots
read the original abstract
Recent model merging methods demonstrate that the parameters of fully-finetuned models specializing in distinct tasks can be combined into one model capable of solving all tasks without retraining. Yet, this success does not transfer well when merging LoRA finetuned models. We study this phenomenon and observe that the weights of LoRA finetuned models showcase a lower degree of alignment compared to their fully-finetuned counterparts. We hypothesize that improving this alignment is key to obtaining better LoRA model merges, and propose KnOTS to address this problem. KnOTS uses the SVD to jointly transform the weights of different LoRA models into an aligned space, where existing merging methods can be applied. In addition, we introduce a new benchmark that explicitly evaluates whether merged models are general models. Notably, KnOTS consistently improves LoRA merging by up to 4.3% across several vision and language benchmarks, including our new setting. We release our code at: https://github.com/gstoica27/KnOTS.
Forward citations
Cited by 12 Pith papers
-
Model Merging as Probabilistic Inference in Fine-Tuning Parameter Space
Model merging is cast as PoE inference with EBM experts, revealing Gaussian assumptions in prior work and proposing convergent Cauchy experts that improve empirical performance.
-
Compositional Generalization in Autoregressive Models via Logit Composition
Logit composition of autoregressive models is projective under factorized conditionals, preserved under smooth reparameterizations, and maintains length generalization when assumptions hold uniformly.
-
DA-MergeLoRA: Hypernetwork-Based LoRA Merging for Few-Shot Test-Time Domain Adaptation
A hypernetwork generates per-column merging weights to combine source LoRA modules on CLIP, achieving state-of-the-art few-shot test-time domain adaptation.
-
Dynamic Model Merging Made Slim
DiDi-Merging achieves dynamic model merging performance matching or exceeding prior methods while using only 1.24x to 1.4x the parameters of a single fine-tuned model.
-
ACE-LoRA: Adaptive Orthogonal Decoupling for Continual Image Editing
ACE-LoRA introduces adaptive orthogonal decoupling and rank-invariant compression for continual image editing in diffusion models, plus the CIE-Bench benchmark.
-
Octopus: History-Free Gradient Orthogonalization for Continual Learning in Multimodal Large Language Models
Octopus introduces history-free gradient orthogonalization in a two-stage finetuning framework to achieve state-of-the-art continual learning results for multimodal LLMs on the UCIT benchmark.
-
Explaining and Breaking the Safety-Helpfulness Ceiling via Preference Dimensional Expansion
MORA breaks the safety-helpfulness ceiling in LLMs by pre-sampling single-reward prompts and rewriting them to incorporate multi-dimensional intents, delivering 5-12.4% gains in sequential alignment and 4.6% overall i...
-
Explaining and Breaking the Safety-Helpfulness Ceiling via Preference Dimensional Expansion
MORA breaks the safety-helpfulness trade-off in LLM alignment by pre-sampling single-reward prompts and rewriting them to expand multi-dimensional reward diversity, yielding 5-12.4% single-preference gains in sequenti...
-
PivotMerge: Bridging Heterogeneous Multimodal Pre-training via Post-Alignment Model Merging
PivotMerge merges post-alignment cross-modal projectors from heterogeneous MLLM pre-training via shared-space filtering and layer-wise weights, beating prior merging baselines on multimodal benchmarks.
-
PivotMerge: Bridging Heterogeneous Multimodal Pre-training via Post-Alignment Model Merging
PivotMerge merges heterogeneous multimodal pre-trained models via shared-space decomposition to filter conflicts and layer-wise weights based on alignment contributions, outperforming baselines on multimodal benchmarks.
-
Model soups need only one ingredient
A single checkpoint, edited by splitting each layer's update with SVD and reweighting the high- and low-energy parts, reaches soup-level OOD robustness without multi-model training.
-
GRIP: In-Parameter Graph Reasoning through Fine-Tuning Large Language Models
An LLM can memorize a knowledge graph into LoRA weights and answer relation/reasoning queries about it without graph context, but the evaluation partly trains on the test task.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.