Anchorless Diversification for Parallel LLM Ideation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Ibrahim, Fares Nabil, Azad, Nafis Saami, Baten, Raiyan Abdul
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917544551514112
author Ibrahim, Fares Nabil
Azad, Nafis Saami
Baten, Raiyan Abdul
author_facet Ibrahim, Fares Nabil
Azad, Nafis Saami
Baten, Raiyan Abdul
contents LLMs are increasingly used to generate candidate-idea pools for creative tasks where broad exploration is valuable. Parallel inference can be attractive in this setting when it broadens the pool while retaining quality and cost efficiency. We study inference-time controls for candidate-pool diversification, asking whether anchorless methods can rival methods that depend on observed seed ideas. Across three creative task families, we compare independent generation and semantic direction stratification with self-, peer-, and representative-anchor baselines, under neutral and population-referential divergent instructions. Population-referential divergence is a strong low-cost baseline, increasing semantic diversity while preserving quality proxies. Semantic direction stratification is stronger: a single planning call organizes generations across broad semantic directions, yielding the best diversity--quality--compute frontier. Anchored regeneration can be strong in final-pool diversity, but its advantage shrinks under full-pipeline token accounting. These results establish practical anchorless baselines for open-ended LLM ideation.
format Preprint
id arxiv_https___arxiv_org_abs_2605_30150
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Anchorless Diversification for Parallel LLM Ideation
Ibrahim, Fares Nabil
Azad, Nafis Saami
Baten, Raiyan Abdul
Artificial Intelligence
LLMs are increasingly used to generate candidate-idea pools for creative tasks where broad exploration is valuable. Parallel inference can be attractive in this setting when it broadens the pool while retaining quality and cost efficiency. We study inference-time controls for candidate-pool diversification, asking whether anchorless methods can rival methods that depend on observed seed ideas. Across three creative task families, we compare independent generation and semantic direction stratification with self-, peer-, and representative-anchor baselines, under neutral and population-referential divergent instructions. Population-referential divergence is a strong low-cost baseline, increasing semantic diversity while preserving quality proxies. Semantic direction stratification is stronger: a single planning call organizes generations across broad semantic directions, yielding the best diversity--quality--compute frontier. Anchored regeneration can be strong in final-pool diversity, but its advantage shrinks under full-pipeline token accounting. These results establish practical anchorless baselines for open-ended LLM ideation.
title Anchorless Diversification for Parallel LLM Ideation
topic Artificial Intelligence
url https://arxiv.org/abs/2605.30150