VibeServe demonstrates that AI agents can synthesize bespoke LLM serving systems end-to-end, remaining competitive with vLLM in standard settings while outperforming it in six non-standard scenarios involving unusual models, workloads, or hardware.
hub
Gated Delta Networks: Improving Mamba2 with Delta Rule
26 Pith papers cite this work. Polarity classification is still indexing.
26
Pith papers citing it
abstract
Linear Transformers have gained attention as efficient alternatives to standard Transformers, but their performance in retrieval and long-context tasks has been limited. To address these limitations, recent work has explored two distinct mechanisms: gating for adaptive memory control and the delta update rule for precise memory modifications. We observe that these mechanisms are complementary: gating enables rapid memory erasure while the delta rule facilitates targeted updates. Building on this insight, we introduce the gated delta rule and develop a parallel training algorithm optimized for modern hardware. Our proposed architecture, Gated DeltaNet, consistently surpasses existing models like Mamba2 and DeltaNet across multiple benchmarks, including language modeling, common-sense reasoning, in-context retrieval, length extrapolation, and long-context understanding. We further enhance performance by developing hybrid architectures that combine Gated DeltaNet layers with sliding window attention or Mamba2 layers, achieving both improved training efficiency and superior task performance.
hub tools
representative citing papers
Chem-GMNet uses sphere-native embeddings, DualSKA attention, and SH-FFN layers to match or beat ChemBERTa-2 on MoleculeNet tasks with fewer parameters and sometimes no pretraining.
SpikeProphecy decomposes spike-count forecasting performance into temporal fidelity, spatial pattern accuracy, and magnitude-invariant alignment, revealing reproducible brain-region predictability rankings and a sub-Poisson evaluation floor across seven model families on 105 Neuropixels sessions.
Mixture of Layers replaces monolithic transformer blocks with routed thin parallel blocks using hybrid attention that combines a shared softmax block for global context with Gated DeltaNet linear attention in the routed blocks.
SATFormer uses a context-dependent gate for selective reuse of early Transformer representations, improving validation loss and zero-shot accuracy especially on retrieval benchmarks.
Preconditioned delta-rule models with a diagonal curvature approximation improve upon standard DeltaNet, GDN, and KDA by better approximating the test-time regression objective.
Mem3R achieves better long-sequence 3D reconstruction by decoupling tracking and mapping with a hybrid memory of TTT-updated MLP and explicit tokens, reducing model size and trajectory errors.
S0 tuning optimizes initial recurrent states in hybrid models to outperform LoRA with zero inference cost on HumanEval and partial cross-domain transfer.
A 130M-parameter 1-layer GPN achieves FineWeb-Edu perplexity 18.06, within 13% of a 12-layer Transformer++ (16.05) and 18% of a 10-layer GDN (15.34).
Pruning pretrained MoE models outperforms training from scratch, different compression methods converge after continued pretraining, and combining KD with language modeling loss plus progressive schedules yields a competitive 23A2B model from Qwen3-Next-80A3B.
Spectral Koopman operators let SSMs achieve 100% accuracy on long-gap multi-query associative recall with fixed memory, where pure Mamba fails.
Cubit replaces Transformer attention with Kernel Ridge Regression token mixing and shows potential gains on longer sequences.
Training transformers with KV sparsification during continued pretraining produces representations that admit better post-hoc KV cache compression, improving quality under memory budgets for long-context tasks.
No model can achieve efficiency, compactness, and recall capacity scaling with sequence length at once, as any two imply a strict bound of O(poly(d)/log V) on recallable facts.
FADE adapts per-parameter weight decay rates online via approximate meta-gradient descent to improve controlled forgetting over fixed decay in online tracking and streaming classification.
HyLo upcycles Transformer LLMs into hybrids with MLA and Mamba2/Gated DeltaNet blocks via staged training and distillation, extending context to 2M tokens and outperforming prior upcycled hybrids on long-context benchmarks.
Gist Sparse Attention uses learnable gist compression tokens as both summaries and routing signals, then selectively unfolds relevant raw chunks for fine-grained attention, outperforming compression and sparse-attention baselines on LongBench and RAG tasks at 8x-32x compression.
In-Place TTT adapts LLM MLP projection matrices at test time with a next-token-aligned objective and chunk-wise updates, enabling better long-context performance as a drop-in enhancement.
A 7B hybrid attention-recurrent model outperforms its pure-transformer counterpart on pretraining metrics and scales more efficiently, supported by a proof that hybrids are strictly more expressive than either transformers or linear RNNs.
Titans combine attention for current context with a learnable neural memory for long-term history, achieving better performance and scaling to over 2M-token contexts on language, reasoning, genomics, and time-series tasks.
Temporal Operator Attention augments softmax attention with learnable sequence-space operators for signed temporal mixing and uses stochastic regularization to enable practical training, yielding consistent gains on time series benchmarks.
Mela is a Transformer variant with a dual-frequency Hierarchical Memory Module and MemStack that performs test-time memory consolidation, outperforming baselines on long contexts.
Irminsul recovers up to 83% of prompt tokens above exact-prefix matching and delivers 63% prefill energy savings per cache hit on MLA-MoE models by content-hashing CDC chunks and applying closed-form kr correction.
Reasoning augmentation extends the difficulty range for both architectures, but hybrid models stay robust longer than transformers as sequential dependence increases in state-based recall tasks.
citing papers explorer
-
VibeServe: Can AI Agents Build Bespoke LLM Serving Systems?
VibeServe demonstrates that AI agents can synthesize bespoke LLM serving systems end-to-end, remaining competitive with vLLM in standard settings while outperforming it in six non-standard scenarios involving unusual models, workloads, or hardware.