AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

Post-training Large Language Models for Diverse High-Quality Responses

Yilei Chen, Souradip Chakraborty, Lorenz Wolf, Ioannis Paschalidis, Aldo Pacchiano

ICLR 2026 · Large language models › Language, safety and interpretability · approach: LLM pruning and scaling data · 14 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add DPP determinant diversity term (DQO)PPO · Dollydistinct-4: 0.64 -> 0.69Improved
add DPP determinant diversity term (DQO)PPO · Dollypass@10: 8.39 -> 8.74Improved
add DPP determinant diversity term (DQO)GRPO · GSM8Kdistinct-4: 0.32 -> 0.42Improved
add DPP determinant diversity term (DQO)GRPO · GSM8Kpass@1: 76.8 -> 76.3No clear effect
add DPP determinant diversity term (DQO)GRPO · GSM8Kpass@10: 87.9 -> 91.2Improved
use DQO instead of GRPO-likelihoodGRPO-likelihood · GSM8Kpass@1: 50.9 -> 76.3Improved
use DQO instead of GRPO-entropyGRPO-entropy · Dollypass@1: 4.71 -> 5.92Improved
use DQO instead of GRPO-entropyGRPO-entropy · GSM8Kpass@10: 92.6 -> 91.2Hurt
increase diversity weight alpha from 1.0 to 2.0 (ablation)DQO · Dollydistinct-4 and pass@1: distinct-4 0.69 -> 0.82, pass@1 5.92 -> 5.42Mixed
increase group size k from 2 to 8 (ablation)DQO · Dollydistinct-4: 0.62 -> 0.79Improved
replace pairwise distance with determinant diversity (ablation)DQO · city recommendationdiversity of recommended citiesImproved
use outcome reward instead of reward model with diversity objectiveDQO · GSM8Kreward hackingHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy