arXiv:2602. 06136v2 Announce Type: replace Abstract: Test-time adaptation (TTA) offers a compelling remedy for machine learning (ML) models that degrade under domain shifts, improving generalisation on-the-fly with only unlabelled samples.
LEAP is a length prediction system for large language model inference that uses the serving model’s prefill hidden states and a linear readout to estimate output length. It adapts online via recursive least squares, updating the readout with completed requests without costly retraining or gradient computation. Experiments show LEAP maintains strong ranking performance across domain shifts, improves Spearman correlation by 64.7% under large shifts, and reduces per-token latency by 15.9–40.8% while increasing throughput and drastically cutting predictor parameters.
arXiv:2606. 04145v1 Announce Type: cross Abstract: Cloud LLM fine-tuning platforms increasingly serve RLHF workloads, where a learned reward model is optimized as a proxy for human quality.
arXiv:2607. 04919v1 Announce Type: new Abstract: Deploying a time series foundation model requires GPU infrastructure, engineering overhead, and carries no guarantee of improvement over XGBoost.
The paper benchmarks two frozen forecasting models, Reverso and Chronos‑Bolt, as feature extractors for time‑series classification on the MONSTER dataset. Using 1%, 10%, and 100% of training labels, the authors compare the models’ frozen representations to four established classifiers and to random‑weight versions of the same models. Results show that pretraining yields higher accuracy than random weights on five of six datasets, but the frozen representations never outperform the best non‑pretrained baseline, performing worse in 15 of 17 settings and tying only twice.
arXiv:2606. 17660v1 Announce Type: cross Abstract: Fine-tuning large language models (LLMs) is compute-intensive and error-prone: model performance depends sensitively on data quality and hyperparameter choices, and na\"ive runs can even degrade model performance.
arXiv:2606. 18431v1 Announce Type: new Abstract: LLM serving exhibits extreme length variability, making size-based scheduling difficult in practice.
The paper introduces JIL, an attack that manipulates length predictions used by large language model (LLM) schedulers to gain higher priority and reduce completion time. By optimizing an adversarial suffix, JIL causes lightweight output-length probes to underestimate request lengths, leading to up to an 83.4 % reduction in predicted length and an average 1.53× faster completion in end‑to‑end serving experiments. The study also evaluates defenses, showing that grouping predictions into coarse intervals can mitigate JIL’s advantage and protect benign requests.
The study investigates whether learned scoring functions can outperform hand‑tuned heuristics in Kubernetes node‑selection. Two models—a Random Forest on engineered features and a graph neural network on job dependency graphs—were trained on a large production trace; both achieved modest regression gains (R²≈0.042) but lagged behind a simple free‑CPU heuristic in Top‑1 ranking accuracy (65‑66% vs. 74‑84%). The authors attribute this gap to objective mismatch, noting that pointwise regression rather than a ranking‑specific loss likely limits performance.
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.
arXiv:2606. 07565v1 Announce Type: new Abstract: Intelligent scaling of microservices in cloud platforms is crucial for mitigating escalating compute costs while avoiding service disruptions.
arXiv:2610.27895v1 Announce Type: cross Abstract: Direct preference optimisation (DPO) is widely used for language model post-training, but its performance depends strongly on the choice of its hyper...