The paper presents a block‑wise differentiable Sinkhorn attention mechanism designed for long‑context balanced entropic optimal transport on TPU hardware. By stopping a $T$‑step Sinkhorn solve and unrolling a short refinement tail, the authors derive an exact surrogate gradient that achieves efficient block‑wise cost and memory usage. Experimental results on synthetic masked problems and a Pfam protein‑family screen demonstrate high numerical accuracy and sustained throughput on TPU v6e‑8, with notable improvements in reconstruction and sparse cross‑entropy metrics.
arXiv:2607. 24741v2 Announce Type: replace-cross Abstract: Positive two-marginal entropic optimal transport is solved by a nonlinear, positive, order-preserving, homogeneous Sinkhorn map.
arXiv:2607. 20594v1 Announce Type: cross Abstract: When does a weight-tied looped transformer -- one block applied T times -- implement an actual algorithm?
arXiv:2610.15829v2 Announce Type: replace Abstract: Hard-margin SVM training is the distance problem between two convex hulls, which the Triangle Algorithm (TA) solves with a certified duality gap. W...
arXiv:2608. 06762v1 Announce Type: new Abstract: Bisimulation metrics quantify behavioral similarity in Markov decision processes, but their Wasserstein fixed-point operator updates every state pair and incurs quadratic pairwise work.
arXiv:2608.30427v1 Announce Type: cross Abstract: Speculative decoding speeds up generation with an efficient draft model (drafter) that proposes tokens for a target model to verify in one pass, pres...
Gated-delta linear attention layers such as Gated DeltaNet (GDN), Kimi Delta Attention (KDA), and GDN-2 are tuned for GPUs, where small tensor-core tiles and fast vector registers make per-channel dec...
arXiv:2608. 05863v1 Announce Type: new Abstract: Modern models no longer keep a plain KV cache: latent caches, learned sparse selectors and recurrent states each carry the model's memory in a different form, and each fails differently under compression.
arXiv:2607. 14568v1 Announce Type: cross Abstract: A companion study ran a 35B mixture-of-experts model on a 2011 NVIDIA Tesla C2075 (Fermi, sm_20, 6GB) as a GPU-prefill/CPU-decode hybrid, because the 4-bit model did not fit in device memory (arXiv:2606.
arXiv:2611.06828v1 Announce Type: new Abstract: Gated-delta linear attention layers such as Gated DeltaNet (GDN), Kimi Delta Attention (KDA), and GDN-2 are tuned for GPUs, where small tensor-core til...
Page-EntroKV is a new KV‑cache eviction framework designed for grouped‑query attention (GQA) in long‑context autoregressive language models. It pools token importance scores across query heads within each physical group using sink‑isolated Renyi‑2 entropy, then projects these pooled scores onto page frames and performs eviction at the hardware tuple level. The authors prove theoretical guarantees on union overhead, budget preservation, and needle‑retention, and demonstrate on a Qwen2.5‑1.5B‑Instruct pilot that Page‑EntroKV achieves exact union overhead control, eliminates sink masquerade, and retains 100% of critical tokens compared to mean‑pooling baselines.
arXiv:2610. 41218v1 Announce Type: new Abstract: Why does length generalization (LG) depend so sharply on the positional encoding (PE)?
arXiv:2609.01129v1 Announce Type: new Abstract: We identify a recurrent algebraic regularity in Transformer attention: a sparse subset of effective OV operators $T=OV^\top$ nearly closes under compos...