AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

FuseFSS: Efficient Secure LLM Inference with Function Secret Sharing

Yuhan Ma, Yong Li, Stefan Schmid

ICML 2026 · Architectures and efficiency › RNNs and normalization · approach: Differentiable optimization layers · 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
replace Sigma's hand-written nonlinear protocols with FuseFSS compiled two-call gatesFuseFSS · BERT/GPT models, seq 128online latency: BERT-base 1613.80 -> 1149.50 ms; 19-33% reductionImproved
use FuseFSS compiled gatesFuseFSS · BERT/GPT models, seq 128key size and key generation time: BERT-base key size 18.076 -> 13.678 GB; keygen 1.32 -> 1.08 sImproved
use FuseFSS instead of Sigma on fixed-point inferenceFuseFSS · GLUE and LAMBADAaccuracy: BERT-base SST-2 89.33 -> 89.33; GPT-2 LAMBADA 60.59 -> 60.90No clear effect
use FuseFSS compiled path for nExp and reciprocal in softmaxFuseFSS · BERT-base softmax, seq 128time (ms): 223 -> 49Improved
FuseFSS does not speed up max-reduction in softmax muchFuseFSS · BERT-base softmax, seq 128time (ms): 93 -> 77No clear effect
increase sequence length for BERT-base (ablation)FuseFSS · BERT-base, seq 32 to 512online time reduction: 15.1% at 32 -> 30.8% at 512Improved
pad compiled instances to mask-independent public shape (required for security) (ablation)FuseFSS GELU gate · BERT-base GELU gateper-gate time (ms): 0.264 -> 1.014Hurt
cache compiled gate program across layers (ablation)FuseFSS · BERT-base, 12 GELU callstotal init time (ms): 10.389 -> 0.903Improved
use FuseFSS on LLaMA-family modelsFuseFSS · LLaMA-7B, seq 16 to 64end-to-end time speedup: 1.09x at seq 16 -> 1.22x at seq 64Improved
use FuseFSS compiled activation gatesFuseFSS · GELU/SiLU gate microbenchmark L=128per-gate speedup and key size: 1.64-2.46x speedup; 4.96-6.25x smaller keysImproved

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