pith. sign in

arxiv: 2503.14456 · v2 · pith:D4UEVNNCnew · submitted 2025-03-18 · 💻 cs.CL · cs.AI· cs.LG

RWKV-7 "Goose" with Expressive Dynamic State Evolution

classification 💻 cs.CL cs.AIcs.LG
keywords rwkv-7billionlanguagemodelsconstantdatasetgoosehttps
0
0 comments X
read the original abstract

We present RWKV-7 "Goose", a new sequence modeling architecture with constant memory usage and constant inference time per token. Despite being trained on dramatically fewer tokens than other top models, our 2.9 billion parameter language model achieves a new 3B SoTA on multilingual tasks and matches the current 3B SoTA on English language downstream performance. RWKV-7 introduces a newly generalized formulation of the delta rule with vector-valued gating and in-context learning rates, as well as a relaxed value replacement rule. We show that RWKV-7 can perform state tracking and recognize all regular languages, while retaining parallelizability of training. This exceeds the capabilities of Transformers under standard complexity conjectures, which are limited to $\mathsf{TC}^0$. To demonstrate RWKV-7's language modeling capability, we also present an extended open source 3.1 trillion token multilingual corpus, and train four RWKV-7 models ranging from 0.19 billion to 2.9 billion parameters on this dataset. To foster openness, reproduction, and adoption, we release our models and dataset component listing at https://huggingface.co/RWKV, and our training and inference code at https://github.com/RWKV/RWKV-LM all under the Apache 2.0 License.

This paper has not been read by Pith yet.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 29 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Test-Time Training with KV Binding Is Secretly Linear Attention

    cs.LG 2026-02 conditional novelty 8.0

    Test-time training with KV binding reduces to learned linear attention.

  2. One More Time: Revisiting Neural Quantum States from a Reinforcement Learning Perspective

    cs.LG 2026-07 unverdicted novelty 7.0

    PWO is a trust-region optimizer for autoregressive NQS that improves stability over Adam and stochastic reconfiguration methods while scaling to 1.5B-parameter models on spin systems.

  3. Tapered Language Models

    cs.LG 2026-06 unverdicted novelty 7.0

    Tapered Language Models monotonically decrease MLP width across depth with a cosine schedule, yielding better perplexity and downstream performance than uniform-width baselines across multiple architectures and scales...

  4. An Algebraic View of the Expressivity of Recurrent Language Models

    cs.FL 2026-06 unverdicted novelty 7.0

    A unified algebraic account reduces RNN expressivity to syntactic monoid division in wreath products and shows diagonal state-space models realize every even-modulus counter under unsigned-integer quantization but non...

  5. Preconditioned DeltaNet: Curvature-aware Sequence Modeling for Linear Recurrences

    cs.LG 2026-04 unverdicted novelty 7.0

    Preconditioned delta-rule models with a diagonal curvature approximation improve upon standard DeltaNet, GDN, and KDA by better approximating the test-time regression objective.

  6. Learning State-Tracking from Code Using Linear RNNs

    cs.LG 2026-02 unverdicted novelty 7.0

    Linear RNNs track states from REPL code traces of permutations better than Transformers, but non-linear RNNs outperform them in partially observable probabilistic automata.

  7. Selective Rotary Position Embedding

    cs.CL 2025-11 unverdicted novelty 7.0

    Selective RoPE adds input-dependent rotations to generalize RoPE, showing implicit positional structure in softmax attention and improving performance on language modeling, copying, state tracking, and retrieval when ...

  8. MG-RWKV: Multi-Grained Context-Aware RWKV for Temporal Forgery Localization

    cs.CV 2026-07 unverdicted novelty 6.0

    MG-RWKV combines bidirectional RWKV, multi-granularity mixture of experts, and cross-granularity consistency to achieve state-of-the-art temporal forgery localization with linear complexity.

  9. Architecture-Aware Reinforcement Learning Makes Sliding-Window Attention Competitive in Math Reasoning

    cs.AI 2026-06 unverdicted novelty 6.0

    Reinforcement learning after SFT conversion narrows the performance gap between sliding-window attention and full self-attention on math reasoning benchmarks while preserving linear complexity.

  10. Dynamic Short Convolutions Improve Transformers

    cs.LG 2026-06 unverdicted novelty 6.0

    Dynamic short convolutions applied to key/query/value projections and linear layers in Transformers yield consistent performance gains and 1.33-1.60x compute advantages over standard models on language modeling from 1...

  11. Forget Less, Generalize More: Unifying Temporal and Structural Adaptation for Dynamic Graphs

    cs.LG 2026-05 unverdicted novelty 6.0

    DSRD unifies temporal and structural adaptation for dynamic graphs via a single recurrent retentive state with learnable time-sensitivity parameters in the decay kernels.

  12. Universal Time Series Generation with Neural Controlled Differential Equations

    cs.LG 2026-05 unverdicted novelty 6.0

    Proves SLiCEs are universal time-series generators approximating path laws in W_∞ and proposes G-SLiCEs for path-space flow matching with benefits on irregular grids.

  13. Towards Understanding Self-Pretraining for Sequence Classification

    cs.LG 2026-05 unverdicted novelty 6.0

    Self-pretraining improves Transformer sequence classification by enabling learning of proximity-biased attention from positional encodings that label supervision alone cannot easily acquire from random starts.

  14. OSDN: Improving Delta Rule with Provable Online Preconditioning in Linear Attention

    cs.LG 2026-05 unverdicted novelty 6.0

    OSDN adds online diagonal preconditioning to the Delta Rule, preserving chunkwise parallelism while proving super-geometric convergence and delivering 32-39% recall gains at 340M-1.3B scales.

  15. Elastic Attention Cores for Scalable Vision Transformers

    cs.CV 2026-05 unverdicted novelty 6.0

    VECA learns effective visual representations using core-periphery attention where patches interact exclusively via a resolution-invariant set of learned core embeddings, achieving linear O(N) complexity while maintain...

  16. A Single-Layer Model Can Do Language Modeling

    cs.CL 2026-05 unverdicted novelty 6.0

    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).

  17. Cubit: Token Mixer with Kernel Ridge Regression

    cs.LG 2026-05 unverdicted novelty 6.0

    Cubit replaces Transformer attention with Kernel Ridge Regression token mixing and shows potential gains on longer sequences.

  18. The Impossibility Triangle of Long-Context Modeling

    cs.CL 2026-05 unverdicted novelty 6.0

    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.

  19. Optimal Decay Spectra for Linear Recurrences

    cs.LG 2026-04 unverdicted novelty 6.0

    PoST reparameterizes decay spectra in linear recurrences with geometric log-spacing and position-adaptive scaling to achieve O(exp(-cN/log t)) decay, improving zero-shot language modeling and long-context retrieval ac...

  20. Gated KalmaNet: A Fading Memory Layer Through Test-Time Ridge Regression

    cs.LG 2025-11 unverdicted novelty 6.0

    Gated KalmaNet uses exact Kalman gain computation with adaptive gating and Chebyshev iteration to improve SSM performance on long-context tasks over prior approximations like DeltaNet.

  21. Higher-order Linear Attention

    cs.LG 2025-10 unverdicted novelty 6.0

    Higher-order Linear Attention realizes second-order and higher interactions in linear-time causal attention via constant-size state and associative scans.

  22. Kimi Linear: An Expressive, Efficient Attention Architecture

    cs.CL 2025-10 unverdicted novelty 6.0

    Kimi Linear hybridizes linear attention with a new KDA module to beat full attention on tasks while slashing KV cache by 75% and speeding decoding up to 6x.

  23. Cubit: Token Mixer with Kernel Ridge Regression

    cs.LG 2026-05 unverdicted novelty 5.0

    Cubit replaces Transformer's attention with a closed-form Kernel Ridge Regression token mixer and reports larger gains as training sequence length increases.

  24. MDN: Parallelizing Stepwise Momentum for Delta Linear Attention

    cs.LG 2026-05 unverdicted novelty 5.0

    MDN parallelizes stepwise momentum for delta linear attention using geometric reordering and dynamical systems analysis, yielding performance gains over Mamba2 and GDN on 400M and 1.3B models.

  25. FG$^2$-GDN: Enhancing Long-Context Gated Delta Networks with Doubly Fine-Grained Control

    cs.LG 2026-04 unverdicted novelty 5.0

    FG²-GDN replaces the scalar beta in the delta update with a channel-wise vector and decouples key/value scaling to improve recall over prior GDN and KDA models.

  26. Linear-Time and Constant-Memory Text Embeddings Based on Recurrent Language Models

    cs.CL 2026-04 unverdicted novelty 5.0

    Fine-tuned recurrent models like Mamba2 produce competitive text embeddings with linear-time constant-memory inference via vertical chunking, outperforming transformers in memory use.

  27. Nirvana: A Specialized Generalist Model With Task-Aware Memory Mechanism

    cs.LG 2025-10 unverdicted novelty 5.0

    Nirvana adds a task-aware memory trigger and updater to specialized generalist models, achieving strong general benchmark results, lowest perplexity in biomedicine/finance/law, and improved MRI reconstruction fidelity.

  28. When Good Enough Is Optimal: Multiplication-Only Matrix Inversion Approximation for Quantized Gated DeltaNet

    cs.LG 2026-06 unverdicted novelty 4.0

    A multiplication-only truncated Neumann approximation for matrix inversion in quantized Gated DeltaNet linear attention delivers up to 5x kernel speedup and 20% decode overhead reduction while preserving accuracy on Q...

  29. Hybrid Architectures for Language Models: Systematic Analysis and Design Insights

    cs.CL 2025-10 unverdicted novelty 4.0

    This work systematically compares inter-layer and intra-layer hybridization strategies for combining self-attention and Mamba-style state space models, evaluating them on language modeling, downstream tasks, long-cont...