The paper introduces VANE, a new routing strategy for Mixture-of-LoRA-experts that scores experts based on the expected loss reduction of their updates rather than token similarity. By decomposing each expert into a reader and writer, VANE evaluates all reader–writer pairs using a low‑rank compass that predicts the descent direction, enabling efficient top‑k selection with additive gates. Experiments on Llama‑3 models show VANE outperforms twelve PEFT and MoE‑LoRA baselines while using fewer trainable parameters and providing router scores that better reflect expert usefulness.
arXiv:2607. 26052v2 Announce Type: replace Abstract: Mixture-of-Experts (MoE) variants of Low-Rank Adaptation (LoRA) route every token to a fixed number of experts $k$.
The paper introduces READ, a method for composing low‑rank adapters (LoRA) in large language models. By rewriting each adapter into a balanced canonical form and enforcing a one‑directional coupling, READ allows new skills to read but never write into the output subspaces of existing skills, eliminating interference. Experiments on four benchmark suites and two model families show that READ consistently outperforms existing baselines, improving SuperGLUE scores by over twenty points and domain suite scores by more than seven points.
arXiv:2608. 07814v1 Announce Type: cross Abstract: Mixture-of-Experts (MoE) language models deliver high capacity at low per-token compute, but deploying them cheaply requires compressing their many expert weight matrices.
arXiv:2608. 07890v1 Announce Type: cross Abstract: Mixture-of-Experts (MoE) models decouple total parameters from per-token compute, but deployment still requires storing every expert.
arXiv:2608. 08853v1 Announce Type: new Abstract: Sparse Mixture-of-Experts (MoE) routers commonly use the same scores both to select experts and to weight their already-computed outputs.
arXiv:2606. 31413v1 Announce Type: new Abstract: Composing independently trained LoRA adapters into a single large language model is useful for multi-domain adaptation, especially when the original training data cannot be shared.
IntBMoE introduces a block‑conditioned mixture‑of‑experts that decouples participation, execution, and materialization by combining dense expert composition with sparse block execution. Each internal layer uses a lightweight hypernetwork to merge all expert bases into a single composed expert, while a router selects only a few blocks per token, keeping compute and memory costs low. Experiments on image classification, language modeling, and sequential recommendation demonstrate consistent performance gains, and the model is deployed in AMap’s generative recommendation system, improving UVCTR by 2.4% in online A/B tests.
JevSoup is a training‑free framework that separates System One expert routing from System Two execution for low‑rank adaptation (LoRA) models. It selects two experts based solely on input and expert descriptions, retains the leading expert’s update, projects the second onto the orthogonal complement of the first update’s row space, and combines them with equal weights. On 14 PorTAL tasks and three Qwen3 scales, JevSoup improves task‑macro accuracy by up to 1.19 % and sample‑micro accuracy by up to 1.21 % over the strongest evaluated baselines.
LoRA-TSD introduces a new optimizer for low‑rank adaptation (LoRA) that treats each update as a tangent vector on the fixed‑rank matrix manifold and applies a Muon‑style spectral‑norm steepest‑descent step within that tangent space. The method avoids costly full‑matrix operations and offers a retraction that is up to 2.8× cheaper than previous manifold approaches. The authors prove that their surrogate recovers LoRA‑Pro, identify the Riemannian gradient as the natural stationarity measure, and provide the first global convergence guarantees for both LoRA‑Pro and LoRA‑TSD, achieving superior performance across multiple benchmarks with Llama and Qwen models.
ACE introduces a training‑free, calibration‑free framework for adaptive expert skipping in Mixture‑of‑Experts LLMs. It combines a Global Spectral Proxy that estimates global transformation capacity with a Router‑Conditioned Refinement that builds expert‑specific direction prototypes, enabling the model to skip low‑contribution experts while always keeping the top‑1 expert. Offline computation of expert statistics leaves only lightweight table lookups during inference, and experiments on three MoE‑based LLMs show ACE outperforms static and dynamic baselines, especially at high skipping ratios.
arXiv:2601. 13020v2 Announce Type: replace-cross Abstract: Continual instruction tuning (CIT) requires multimodal large language models (MLLMs) to adapt to a stream of tasks without forgetting prior capabilities.
Low-rank adaptation (LoRA) is the standard way to fine-tune large models, yet when its two factors are trained independently, the update ignores the geometry of the low-rank weight change it induces. We introduce LoRA-TSD, an optimizer that treats every LoRA step as a tangent vector of the fixed-rank matrix manifold and takes the spectral-norm steepest-descent step of Muon inside that tangent space, mapping the result back to the factors through a retraction native to the LoRA parametrization.