提出两层架构让26种模型合并策略支持无冲突分布式操作
Conflict-Free Replicated Data Types for Neural Network Model Merging: A Two-Layer Architecture Enabling CRDT-Compliant Model Merging Across 26 Strategies
- 用CRDT层包装任意合并策略，分离贡献管理与实际合并
- 在3个层级测试中均实现强最终一致性，延迟低于0.5毫秒
- 适合需要分布式训练或模型版本同步的团队使用
我们测试了包括权重平均、SLERP、TIES、DARE、Fisher合并和进化方法在内的26种神经网络合并策略，发现它们均不满足冲突自由分布式操作所需的代数性质（交换律、结合律、幂等性）。我们证明这种失败是结构性的：基于归一化的合并无法同时满足所有三个性质。为此，提出两层架构CRDTMergeState：第一层采用OR-Set CRDT语义管理贡献，合并操作为集合并集，天然满足交换律、结合律和幂等性；第二层在规范排序的贡献集上以确定性纯函数应用合并策略，随机性由梅尔克尔根生成。我们证明该设计保证强最终一致性：所有接收相同贡献的副本无论消息顺序如何，都能计算出相同的合并模型。实证验证覆盖三个层级：控制4×4张量（104/104次测试通过）、高达72.4亿参数的生产级模型（208次策略级测试，43,368次层级属性检查，张量分辨率受限），以及100节点在流言传播和分区恢复下的多节点收敛（20种消息顺序），CRDT开销低于0.5毫秒。由于封装透明，下游性能完全一致，经字节级输出验证。参考实现已发布为crdt-merge v0.9.4。
原文摘要 · Abstract (English)
All 26 neural network merge strategies we tested including weight averaging, SLERP, TIES, DARE, Fisher merging, and evolutionary approaches -- fail the algebraic properties (commutativity, associativity, idempotency) required for conflict-free distributed operation. We prove that this failure is structural: normalisation-based merges cannot simultaneously satisfy all three properties. To resolve this, we present a two-layer architecture -- CRDTMergeState -- that wraps any merge strategy in a CRDT-compliant (Conflict-Free Replicated Data Type) layer. Layer 1 manages contributions via OR-Set CRDT semantics, where the merge operation is set union -- trivially commutative, associative, and idempotent. Layer 2 applies merge strategies as deterministic pure functions over a canonically-ordered contribution set, with randomness seeded from the Merkle root. We prove that this separation guarantees Strong Eventual Consistency: all replicas receiving the same contributions compute identical merged models, regardless of message ordering. Empirical validation spans three tiers: controlled 4x4 tensors (104/104 tests pass), production-scale models up to 7.24B parameters (208 strategy-level tests, 43,368 layer-level property checks at capped tensor resolution), and multi-node convergence under gossip and partition healing (100 nodes, 20 orderings), with CRDT overhead below 0.5 ms. Because the wrapper is transparent, downstream performance is identical by construction, confirmed via byte-identical output verification. The reference implementation is available as crdt-merge v0.9.4.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。