paper
Forecasting With LLMs: Improved Generalization Through Feature Steering
paperactiveprovisional
forecasting-with-llms-improved-generalization-through-feature-steering-8d6e4f01·1 events·first seen 2d agoAliases: Forecasting With LLMs: Improved Generalization Through Feature Steering
Co-occurring entities
More like this (12)
Agentic Chain-of-Thought Steering for Efficient and Controllable LLM ReasoningBackdoor Unlearning Generalization: A Path Toward the Removal of Unknown Triggers in LLMsPredicting Future Behaviors in Reasoning Models Enables Better SteeringLearning from the Self-future: On-policy Self-distillation for dLLMsProxy Metrics for LLM ForecastingVisual Verification Enables Inference-time Steering and Autonomous Policy ImprovementContinual LLM Upcycling: A Predictor-Gated Bank-Wise Sparsity Training Recipe for Dense-to-Sparse LLMsExpRL: Exploratory RL for LLM Mid-TrainingLeveraging Audio-LLMs to Filter Speech-to-Speech Training DataPC Layer: Polynomial Weight Preconditioning for Improving LLM Pre-TrainingInSight: Self-Guided Skill Acquisition via Steerable VLAsInSight: Self-Guided Skill Acquisition via Steerable VLAs
Recent events (1)
Feature steering via sparse autoencoders reduces look-ahead bias in LLM forecasting
Researchers apply sparse autoencoders to inspect LLM internal states during forecasting tasks, identifying features associated with time-aware versus look-ahead-biased reasoning. Amplifying time-awareness features causally reduces look-ahead bias while preserving general reasoning performance, whereas directly steering look-ahead-bias features has no effect. The work demonstrates that interpretable temporal features can shift LLMs toward more historically grounded forecasting. This is a mechanistic interpretability result with practical implications for LLM-based prediction systems.