Data and code for: More Agents Is Not Enough: A Budget-Aware Evaluation of Fine-Tuned Multi-Agent LLM Systems for Hate-Speech Target Categorization

Companion deposit for the manuscript "More Agents Is Not Enough: A Budget-Aware Evaluation of Fine-Tuned Multi-Agent LLM Systems for Hate-Speech Target Categorization", submitted to the Journal of Computational and Cognitive Engineering. The study compares an aggregative ensemble of three fine-tuned persona agents and a compositional three-stage pipeline with a monolithic model and a self-consistency control at measured generated-token budgets, on four-way hate-speech target categorization, across five training seeds (42–46), with one Llama-3-8B base adapted with rationales from a Llama-3-70B teacher, under two readouts (normalized label-continuation scores and generated labels). The deposit contains: the canonical train/validation/test split (8,180 items, split seed pinned); per-item predictions and aligned test tables for every arm at every seed; measured inference cost and hardware logs; the dual-readout recomputation and the verdict-stripped re-embedding outputs; the external best-of-N probe (ten samples of the frozen seed-42 monolith at temperature 1.2, its reproduction gate and its trained verifiers); the trained LoRA adapters (rank 16, 41,943,040 trainable parameters) for every arm and seed; and the notebooks, audit scripts and verification scripts that recompute the main results from the stored predictions, without retraining or re-inference. Version 1.1.0 adds supplement_reanalysis.zip, the reanalysis package of the submitted manuscript (an independent recomputation of every reported number, the recovery of the gold labels from the raw HateXplain and Measuring Hate Speech releases, and the written analysis plan), and DATA_DICTIONARY.md, and updates README.md to the submitted manuscript; all other files are unchanged from version 1.0.0. MANIFEST.json lists the MD5 of every file. Built with Meta Llama 3. The adapters are derivative works of Meta Llama 3 and are distributed under the Meta Llama 3 Community License; they are released for moderation research only. Source corpora are redistributed under their original licences with attribution: HateXplain (Mathew et al., 2021; MIT) and the Measuring Hate Speech corpus (Kennedy et al., 2020; Sachdeva et al., 2022; CC BY 4.0).