4arXiv cs.LG (Machine Learning)·
TabPack: Hyperparameter-ensemble MLP method for tabular deep learning with strong out-of-the-box performance
TabPack is a new efficient ensemble method for tabular deep learning that samples and trains many MLPs with different hyperparameters in parallel within a single run, selecting ensemble members on the fly. The approach eliminates the need for precise hyperparameter tuning by requiring only sampling ranges rather than exact values. On medium-to-large public datasets, default TabPack settings match extensively tuned prior methods while running faster than some baselines even on consumer hardware.
Related guides (1)
Related events (8)
Ensembling Tabular Foundation Models: A Diversity Ceiling and a Calibration Trap
This paper benchmarks six ensemble strategies across six tabular foundation models (TFMs) on 153 OpenML classification tasks, finding that ensembling provides minimal gains over the best single TFM. The best ensemble strategy (two-level cascade stacking) achieves only +0.18% accuracy improvement at 253× the compute cost. A key finding is that logistic-regression meta-learner stacking improves accuracy while severely degrading calibration (log-loss), because sharpening class boundaries destroys probability estimates. The authors recommend greedy ensemble selection as the practical default.
Empirical evaluation finds tabular foundation models degrade systematically under distribution shift
A new arXiv preprint evaluates nine tabular foundation models (TFMs) — including TabPFNv2 through v3, TabICL, Mitra, LimiX, and TabFM — on out-of-distribution performance using three real-world datasets from the TableShift benchmark covering label, socioeconomic, and geographic shifts. All evaluated TFMs show systematic performance degradation under distribution shift regardless of pre-training strategy, with shift gaps ranging from 0.003 to 0.060. The study also identifies a scalability gap, noting that high-performing TFMs require computational resources beyond standard deployment infrastructure. Results suggest TFMs inherit the same OOD fragility documented in classical tabular models, raising concerns for high-stakes deployment.
KnowsTFM: Knowledge graph-informed fine-tuning of small tabular foundation models
A new arXiv preprint introduces KnowsTFM, a method for fine-tuning small tabular foundation models (nanoscale TabPFN and TabICL variants) using structural attention priors derived from knowledge graphs and parameter-efficient low-rank updates. The approach targets niche domains with scarce, high-dimensional data shifted from pretraining distributions, showing meaningful gains in specialist settings but marginal gains on general tasks. The paper also reports that continual fine-tuning of frontier tabular models can trigger collapse of pretrained knowledge, a notable failure mode.
Quantifying Hyperparameter Transfer: Embedding Layer Learning Rate as Key Driver of μP Benefits
This paper develops a three-metric framework to quantify hyperparameter transfer quality across model scales, targeting the problem of extrapolating optimal hyperparameters from small to large LLMs. The central empirical finding is that the well-known advantage of Maximal Update Parameterization (μP) over standard parameterization (SP) with AdamW largely reduces to a single factor: the embedding layer learning rate. In SP, the embedding layer acts as a training bottleneck causing instabilities; scaling its learning rate by model width to match μP substantially stabilizes training and improves transfer. The paper also characterizes how weight decay affects scaling law fit quality versus extrapolation robustness in opposite directions.
Distilling Tabular Foundation Models for Structured Health Data
This paper investigates knowledge distillation from tabular foundation models (TFMs) to lightweight student models for healthcare applications. The authors address context leakage in in-context TFMs via stratified out-of-fold teacher labeling, evaluating across 19 healthcare datasets, 6 TFM teachers, and 4 student families. Distilled students retain at least 90% of teacher AUC while running 26× faster on CPU, with preserved calibration and fairness properties. Multi-teacher ensembles do not consistently outperform the best single teacher.
Mach-Mind-4-Flash: 35B MoE agentic model matching 100B-class performance via post-training optimization
Mach-Mind-4-Flash is a 35B-parameter Mixture-of-Experts model with only 3B activated parameters that achieves performance comparable to 100B-class models through post-training techniques alone. The pipeline combines a unified RL/OPD training infrastructure with multi-teacher scheduling, parallel domain-specific RL experts fused via Multi-Teacher On-Policy Distillation (MOPD), and Hybrid Median-length Policy Optimization (HMPO) which compresses reasoning chains 19-46% with minimal accuracy loss. Benchmark results include 92.70 on AIME'26, 82.82 on IFBench, and 75.80 on BFCL-v4, claiming to lead or match models 10-30x its activated size at a fraction of inference cost. The work is notable for demonstrating that post-training optimization can close large gaps in activated parameter count for agentic tasks.
torchtune: PyTorch Native Post-Training Library for LLMs
Meta's PyTorch team introduces torchtune, a PyTorch-native library for post-training LLMs that emphasizes modularity, hackability, and direct access to underlying PyTorch components. The library supports fine-tuning, experimentation, and deployment-oriented workflows across distributed training settings. Benchmarked against popular frameworks Axolotl and Unsloth, torchtune demonstrates competitive performance and memory efficiency while maintaining flexibility for research iteration. The paper presents design principles, model builders, training recipes, and distributed training stack details.
On the Scaling of PEFT: Towards Million Personal Models of Trillion Parameters
This paper reframes parameter-efficient fine-tuning (PEFT) not merely as a cheaper alternative to full fine-tuning, but as a substrate for persistent, instance-specific personal models layered atop shared foundation models. The authors analyze three scaling axes: Scale Up (stronger base models amplifying adapter utility), Scale Down (minimum viable adapter size), and Scale Out (managing millions of concurrent adapted instances). They introduce MinT as an infrastructure reference for adapter identity, versioning, provenance, evaluation, and serving at scale.