supervised-learning

This paper introduces FormatMix, a multi-format training approach that improves LLM consistency across different answer formats by expanding a subset of training items into multiple equivalent formats, showing that format diversity is key to robustness.

Rich Sutton argues that generative AI trained by supervised learning cannot achieve genuine novelty and quality simultaneously, and that true discovery requires a 'vary, evaluate, select' mechanism found in reinforcement learning rather than pure imitation.

This paper formalizes the concept of Bayes-sufficient representations in supervised learning, defining when a representation retains exactly the information needed for Bayes-optimal prediction under a given loss function. It introduces the Bayes quotient as a canonical loss-dependent object and connects the framework to property elicitation, illustrating distinctions between sufficiency, minimality, and excess retained information through experiments.

This paper proposes Q-align DT, a framework that aligns return-to-go with Q-values to improve controllability and performance in offline reinforcement learning, achieving superior results on D4RL benchmarks.

This paper presents an applied evaluation of foundation models for time series forecasting compared to supervised approaches across four operational domains, and proposes a Complexity Router to selectively assign series to the optimal model class for balancing accuracy and inference cost.

This paper proposes goal-conditioned supervised learning (GCSL) as an offline fine-tuning framework for LLMs, which treats feedback as an explicit goal and trains models via supervised learning with a novel goal formulation and natural-language goal representations. Evaluated on non-toxic generation, code generation, and recommendation, it outperforms standard offline baselines.

This paper investigates whether shallow neural network agents can master the card game Schnapsen using reinforcement learning, outperforming a supervised imitation baseline and achieving competitive results against a strong search-based opponent.

This paper argues that exploration is essential for all learning systems, including supervised learning, and proposes a framework for generalized exploration to drive open-ended learning towards general intelligence.

This paper introduces a unified geometric framework showing that weighted InfoNCE objectives can be interpreted as Distance Geometry Problems, providing exact characterizations of optimal embeddings for supervised and weakly supervised contrastive learning methods and revealing when such embeddings are geometrically realizable, degenerate, or inconsistent.

This paper introduces UniSD, a unified self-distillation framework for adapting large language models that integrates mechanisms for supervision reliability, representation alignment, and training stability. Experimental results show that UniSD improves performance over base models and existing baselines across multiple benchmarks.