Nov 6, 2023 · A multi-reward conditioned self-rationalization algorithm that optimizes multiple distinct properties like plausibility, diversity and consistency.
MaRio (Multi-rewArd RatIOnalization) is a method that tailors small-sized LMs (< 1B parameters) to be strong rationalizers, in terms of both improved ...
Code and Dataset for preprint titled "Tailoring Self-Rationalizers with Multi-Reward Distillation" - INK-USC/RationaleMultiRewardDistillation.
Large language models (LMs) are capable of generating free-text rationales to aid question answering. However, prior work 1) suggests that useful self-.
Nov 5, 2023 · A multi-reward conditioned self-rationalization algorithm that optimizes multiple distinct properties like plausibility, diversity and consistency.
May 23, 2024 · This paper presents a method called MaRio (Multi-reward Rationalization) that enables smaller-scale language models (around 200x smaller than ...
People also search for
Tailoring Self-Rationalizers with Multi-Reward Distillation Sahana Ramnath ... I2D2: Inductive Knowledge Distillation with Neurologic and Self-Imitation
Inference-time policy adapters (ipa): Tailoring extreme-scale lms without fine-tuning ... Tailoring self-rationalizers with multi-reward distillation. S Ramnath, ...
Tailoring Self-Rationalizers with Multi-Reward Distillation · JARVIS-1: Open-World Multi-task Agents with Memory-Augmented Multimodal Language Models · S-LoRA: ...
Tailoring Self-Rationalizers with Multi-Reward Distillation. S Ramnath, B ... Resprompt: Residual Connection Prompting Advances Multi-Step Reasoning in Large ...