AI Paper Atlas

Research map / Federated learning and privacy / Personalized and heterogeneous FL

FedSDR: Federated Self-Distillation with Rectification

Ziheng Ren, Zhanming Shen, Hao Wang, Ning Liu, You Song

ICML 2026 · Federated learning and privacy › Personalized and heterogeneous FL · approach: Non-IID federated learning · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add FedSD self-distillation to FedAvgFedAvg · Dolly-15K Overall ScoreOverall Score: 71.21 -> 74.54Improved
add FedSD to FedAvgMFedAvgM · Dolly-15K Overall ScoreOverall Score: 68.18 -> 73.34Improved
add FedSD to FedYogiFedYogi · Dolly-15K MMLUMMLU: 29.92 -> 39.32Improved
add FedSD to FedProxFedProx · Dolly-15K DROPDROP: 37.50 -> 36.68Hurt
add FedSD to FedAvg under high heterogeneityFedAvg · Dolly-15K alpha=0.1Overall: 53.18 -> 63.26Improved
self-distill client data (rewrite with model)client data · Dolly-15K inter-client divergenceJS Divergence: 0.4074 -> 0.3611Improved
self-distill client dataclient data · Dolly-15K inter-client similarityTF-IDF cosine similarity: 0.6362 -> 0.7064Improved
add dual-stream rectification (LoRA-R, selective aggregation) to FedSD (ablation)FedSD · Dolly-15K extended ablationOverall: 69.05 -> 77.38Improved
upload only LoRA-R instead of both adapters (selective aggregation) (ablation)Dual-LoRA(FedDPA) · Dolly-15K extended ablationOverall: 68.77 -> 77.38Improved
add FedSDR rectification versus FedSD on faithfulness (ablation)FedSD · Dolly-15K extended ablationFaithfulness: 66.67 -> 74.33Improved
scale clients from 8 to 40 with FedSDRFedSDR · Dolly-15K number of clientsOverall: 75.24 -> 77.70Improved
use FedSDR instead of FedAvg on Qwen2.5-7B-Instruct mixed dataFedAvg · GSM8K (mixed 5-domain FL)accuracy: 66.79 -> 78.32Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy