Paper on LibFHE, a Numba/CUDA-Python GPU FHE library for non-RNS CKKS-BGV, because C++ was too mainstream.

Non-RNS CKKS-BGV on GPUs, because apparently someone missed the usual RNS memo and still got competitive performance.

Walsh LUTs on lazy CKKS bits speed up AES transciphering, because apparently XORs deserved a detour.

Non-RNS CKKS-BGV on GPUs, apparently via Numba, and the paper claims C++-library-level performance without the C++.

Paper on prefix scans for noncommutative monoids in bit-reversed HE slots, cutting rotations to the theoretical minimum.

Can SNLP cut FHE Transformer depth? This paper says yes, with fewer bootstraps and only modest perplexity loss.

Generalizes CKKS key-switching decomposition to many inputs/keys, so batched rotations/relinearizations get less noise and less waste.

Can you keep dense retrieval useful after shredding embeddings into public prefixes and private CKKS shards? SHARD says maybe.

TGHE canonizes ego-graphs into shared CKKS packs for encrypted GNN inference, because bespoke graph HE was still too bespoke.

How do you get FHE LLM serving out of the latency swamp? Split linear ops to CKKS, keep the messy bits in a TEE.

Paper on CKKS packing for private inference, using fragment encoding to cut rotations and ciphertexts. Packing, but make it baroque.

WHET is an architecture-aware FHE accelerator paper that cuts memory traffic and bootstrapping latency, because DRAM was the real bottlen...