@inproceedings{anantha-etal-2026-nanoflux,
    title = "{N}ano{F}lux: Adversarial Dual-{LLM} Evaluation and Distillation for Multi-Domain Reasoning",
    author = "Anantha, Raviteja  and
      Hor, Soheil  and
      Antoniu, Teodor Nicola  and
      Price, Layne C",
    editor = "Mille, Simon  and
      Gehrmann, Sebastian  and
      Schmidtov{\'a}, Patr{\'i}cia  and
      Du{\v{s}}ek, Ond{\v{r}}ej  and
      Fadaee, Marzieh  and
      Lo, Kyle  and
      Santus, Enrico  and
      Stanovsky, Gabriel",
    booktitle = "Proceedings of the Fifth Workshop on Generation, Evaluation and Metrics ({GEM})",
    month = jul,
    year = "2026",
    address = "San Diego, California, USA",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-acl-workshops/2026.gem-main.27/",
    pages = "253--270",
    ISBN = "979-8-89176-423-1",
    abstract = "We present NanoFlux, a novel adversarial framework for generating targeted training data to improve LLM reasoning, where adversarially-generated datasets of $\leq 200$ examples outperform conventional fine-tuning approaches. The framework employs a competitive dynamic between models alternating as Attacker and Defender, supervised by a tool-augmented Judge, synthesizing multi-step questions with explanatory annotations. Fine-tuning a 4B-parameter model on NanoFlux-generated data yields performance gains across diverse domains compared to full-benchmark fine-tuning: +5.9{\%} on mathematical reasoning, +3.6{\%} on scientific reasoning, and +16.6{\%} on medical reasoning, while reducing computational requirements by 3-14{\texttimes}. Ablation studies reveal a non-monotonic relationship between dataset characteristics and model performance, uncovering domain-specific optimal points for question complexity and reasoning quality. NanoFlux automates training data generation through embedding-based novelty filtering, tool-augmented evaluation, and multi-hop reasoning, pointing to the value of small, targeted training datasets."
}