Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces TwistedMerge, a novel framework for model merging that addresses the challenges of achieving global alignment among independently trained models. By formulating the merging process as a finite descent problem and employing a conservative certification pipeline, the authors ensure that only reliable merges are certified, while others are abstained from, thereby preventing degradation in model performance. Key results demonstrate that naive averaging methods can lead to instability, while the proposed approach maintains robustness against noise and misalignment, ultimately validating descent theory as a viable framework for model certification and abstention.
Cycle-consistent synchronization reveals that naive factor averaging can destabilize model merging, while TwistedMerge provides a robust certification framework that abstains when necessary.
Model merging combines independently trained or fine-tuned models, but pairwise alignability does not imply globally consistent alignment. We formulate merging as a finite descent problem in which checkpoints are local objects, alignment maps are transitions, and cycle products are residuals. TwistedMerge is a conservative certification pipeline that separates fixed-chart averaging, synchronization-removable gauge inconsistency, a certified central obstruction on a specified comparison complex, and nonabelian holonomy. A residual is promoted to a cohomology class only after inverse-consistency, coefficient-identification, centrality, and closure tests; otherwise the method abstains and returns an ordinary or synchronized fallback. We prove a constant-edge no-go result, frozen-complex three-way and predeclared-family error-control theorems, and a refinement test for comparison-complex sensitivity. A planted neural alignment defect is removed by cycle-consistent synchronization, showing that a nonzero cycle score alone is not a higher obstruction. Controlled central systems recover the predicted non-coboundary and projective-rank behavior, while noisy estimates move from certification to abstention without false lifts on the tested controls. A trained low-rank-adapter audit shows that naive factor averaging depends on the chosen GLr representative, whereas global factor synchronization and dense-delta SVD are stable. On natural checkpoint collections, cycle residuals do not predict merge degradation and no natural central or period-index class is certified. The results position descent theory as a falsifiable certification and abstention framework.