arXiv:2605. 29183v2 Announce Type: replace-cross Abstract: As machine learning(ML) systems evolve to continual adaptation, each re-training cycle uses compute, annotation, and energy.
arXiv:2609.06806v1 Announce Type: new Abstract: Training strategy, namely whether to retrain from scratch or fine-tune from the previous checkpoint, is an overlooked decision variable in active learn...
DUET is a token‑budget allocation method for reinforcement learning with verifiable rewards that jointly optimizes which prompts receive rollouts and how long each rollout runs. By using a lightweight surrogate to assign rollouts and a marker‑gated abort rule to stop them, DUET improves reasoning quality and reduces training time compared to full‑budget GRPO and other baseline methods. Experiments on Qwen3‑1.7B and other LLMs show that DUET outperforms baselines even with only 50% of the token budget, achieving up to 2.51× speedup while maintaining or improving performance.
The paper investigates when online adaptation benefits edge time‑series forecasting under distribution drift, using a leakage‑free streaming protocol on six public multivariate datasets. It shows that the warmup budget for static baselines and the choice of learning rate can bias perceived adaptation gains, and that a validation‑only procedure selecting warmup and optimizer rates yields Adam outperforming SGD with momentum in most settings. The study also examines accuracy versus adaptation‑state memory and per‑update latency for different adaptation strategies, highlighting parameter‑efficient variants that are nondominated on the memory axis.
arXiv:2608. 19488v1 Announce Type: new Abstract: Production machine learning systems degrade under concept drift, yet practitioners have little principled guidance on when to retrain.
MentorPulse introduces a method for refreshing cross‑model latent guidance during long‑form generation. By compressing mentor states into a capped slot memory and updating it incrementally as new tokens are generated, the approach keeps the guidance signal fresh without resetting the student’s KV cache. Across thirteen datasets, MentorPulse closes more than half of the mentor‑student performance gap, outperforming existing techniques such as C2C, T2T, and LoRA, especially on long outputs.
arXiv:2606. 13370v1 Announce Type: new Abstract: This study examines training dynamics in a small Llama-style language model trained under a fixed, compute-constrained token budget.
arXiv:2607. 00958v1 Announce Type: new Abstract: Time series are central to modern data mining applications, from industrial telemetry and server metrics to finance and physiology, yet time-series self-supervised learning often depends on view and augmentation choices that encode domain-specific invariances.
The study investigates whether switching optimizers between pretraining and subsequent training stages of large language models (LLMs) incurs a performance penalty. By pretraining 1.2B Llama‑3 models with varying proportions of AdamW and Muon updates, and interpolating between the two during continued pretraining and supervised fine‑tuning (SFT), the authors find little evidence of a meaningful optimizer mismatch penalty. The results show that final performance is largely determined by the independent effects of the pretraining and post‑training optimizers, with downstream performance generally improving as the proportion of Muon updates during SFT increases, and that Muon remains competitive with AdamW even for task‑general SFT on larger models.
arXiv:2606. 25068v1 Announce Type: new Abstract: Online time-series forecasters receive labels only after horizon-dependent delays, while every adaptation step spends limited compute.
arXiv:2607. 04969v1 Announce Type: new Abstract: The training paradigm of large language models has shifted from traditional one-pass training to multi-epoch training, as reasonable reuse of limited high-quality data can improve both model performance and sample efficiency.
The study investigates whether a curated same-family neural network experiment can guide large language model (LLM)-based improvements for a low-performing target model under equal generation and evaluation budgets. Using TuneNNGen, an extension of NNGPT, the authors compare source-guided generation with target-only generation on CIFAR-10, SVHN, Imagenette, and CIFAR-100 datasets, achieving significant accuracy gains across these benchmarks. The results demonstrate that the benefits depend on source-target compatibility and LLM adaptation, rather than merely on stored source accuracy.
arXiv:2609.36655v1 Announce Type: new Abstract: Continual test-time adaptation (CTTA) adapts a source model to an unlabeled test stream whose distribution may change over time. Existing TTA methods o...