DepthGraft: Structural Regularization through Hierarchical Cross-Layer KV Reconstruction

Xiaohan Qin, Xiangdong Zhang, Yu Wang, Huaijin Wu, Zhuo Xia, Yebin Yang, Junchi Yan

DepthGraft: Structural Regularization through Hierarchical Cross-Layer KV Reconstruction: 15 upvotes on Hugging Face Daily Papers, #13 of 39 papers on 2026-11-11. Day-by-day upvote history.

Upvotes over time

It had 12 upvotes by its first day, 15 upvotes by the next day, 15 upvotes by day 3, 15 upvotes by day 7, and has 15 now. It ranked #43 of 133 papers in its week and #129 of 340 in its month. Its code repository has 23 GitHub stars.

Scaling depth is a key route to stronger large language models, but Pre-LN architectures often suffer from depth-wise information dilution, limiting the effective use of deep layers. While cross-layer connectivity has proven effective at mitigating this problem, existing designs usually introduce additional computation or memory access. We revisit cross-layer KV sharing as an efficient form of such connectivity. Beyond cache compression, each KV reuse edge creates a direct gradient path from deep layers to source KV projections, turning KV sharing into an implicit structural regularizer for KV representation geometry. We provide a systematic analysis of this mechanism, showing that proper sharing topologies sharpen key retrieval directions while enriching the value content subspace, whereas existing adjacent or global-anchor designs do not fully exploit this effect. Motivated by this, we propose DepthGraft, a hierarchical KV-sharing architecture that reconstructs deep-layer KV from stride-aligned shallow-middle source pairs. DepthGraft distributes non-local gradient flow across source layers, while preserving prefill early completion and reducing KV cache by roughly one third. Experiments on dense and MoE models from 0.5B to 16B parameters show consistent downstream improvements under both GQA and MLA. Notably, on the 16B MoE model, DepthGraft achieves a 7.0% absolute improvement on MMLU-Pro, along with gains of 6.6% on CMMLU and 6.4% on C3. These results validate the potential of DepthGraft as a scalable design for large-scale pretraining. The project page is available at https://github.com/jianke0604/DepthGraft.

Paper page on Hugging Face · arXiv

Data: hysts-bot-data/daily-papers-stats and the Daily Papers API. Open data: tardellirs/paper-pulse-data. Sister project: Model Pulse, the download history of every model on the Hub.