Papers2

#Calibration

Blockwise Advantage Estimation for Multi-Objective RL with Verifiable Rewards

Kirill Pavlenko, Alexander Golubev et al.Feb 10arXiv

The paper fixes a common mistake in training language models for multi-part tasks: giving the same reward signal to every token, even when different text parts aim at different goals.

#Blockwise Advantage Estimation#Outcome-Conditioned Baseline#Group Relative Policy Optimization

QuCo-RAG: Quantifying Uncertainty from the Pre-training Corpus for Dynamic Retrieval-Augmented Generation

Intermediate

Dehai Min, Kailin Zhang et al.Dec 22arXiv

QuCo-RAG is a new way to decide when an AI should look things up while it writes, using facts from its training data instead of its own shaky confidence.

#Dynamic RAG#Retrieval-Augmented Generation#Uncertainty Quantification