{
  "abstract": "Decode-time alignment methods steer a frozen language model by scoring candidate continuations with an external reward and selecting the maximiser. We argue that this shared design is a single degenerate point in a much larger space. We introduce Swiss-Knife, a framework for externalised multi-objective alignment in which the alignment specification is a first-class runtime object: hot-swappable scoring blades, a batch normaliser, a pairwise aggregation operator, and a selection rule. Six axioms characterise the admissible aggregation operators, and we prove a representation theorem: every operator satisfying them has the form $R_i = \\sum_{j \\neq i} g((μ_i - μ_j)/s(σ_i,σ_j))$, a two-parameter family containing probit and logistic comparison rules and pointwise argmax as named coordinates. Within it, pairwise aggregation is Lipschitz-stable under adversarial reward contamination while argmax is not, and Candidate-Batch Normalization (CBN) makes the weight simplex invariant to the rescalings under which reward models are only ever identified. Our reference instantiation pairs DPO-LoRA blades with an uncertainty-aware pairwise tournament. Sweeping the helpfulness/honesty/harmlessness simplex, it attains the best balanced frontier of six decode-time methods (harmonic $F_1$ 0.797 vs. 0.750 for the strongest baseline, $p < 10^{-14}$) with the lowest refusal rate and highest helpfulness of any arm, and reconfigures its objectives in 0.05 ms with no gradient computation. Ablating CBN costs 0.217 $F_1$, and the metadata confirms the predicted mechanism: the lowest-variance blade retains 9% of its nominal 33% influence, and the collapse follows the predicted ordering.",
  "abstract_shown": true,
  "affiliations": [],
  "article_type": null,
  "arxiv": "2609.22226",
  "author_total": 8,
  "authors": [
    {
      "name": "Agnibh Karmakar"
    },
    {
      "name": "Mayur Parvatikar"
    },
    {
      "name": "Shreyash Dhoot"
    },
    {
      "name": "Amit Dhanda"
    },
    {
      "name": "Aman Chadha"
    },
    {
      "name": "Kapil Wanaskar"
    },
    {
      "name": "Vinija Jain"
    },
    {
      "name": "Amitava Das"
    }
  ],
  "bibtex": "https://research.draconitic.com/paper/4c78a1k6kn2f2.bib",
  "cited_by": 0,
  "doi": "10.48550/arxiv.2609.22226",
  "fulltext_url": null,
  "id": "4c78a1k6kn2f2",
  "issns": [],
  "landing_url": "https://arxiv.org/abs/2609.22226",
  "language": null,
  "licence": "LicenseRef-arXiv-nonexclusive-1.0",
  "licence_label": "arXiv licence",
  "licence_source": "article",
  "licence_url": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
  "metadata_licence": "CC0-1.0",
  "native_subjects": [
    [
      "arxiv",
      "cs.CL"
    ],
    [
      "arxiv",
      "cs.LG"
    ],
    [
      "arxiv",
      "stat.ML"
    ],
    [
      "openalex",
      "1702|Topic Modeling"
    ]
  ],
  "openalex": "W7214016046",
  "pdf_url": "https://arxiv.org/pdf/2609.22226",
  "pmcid": null,
  "pmid": null,
  "published": "2026-09-03",
  "publisher": null,
  "ris": "https://research.draconitic.com/paper/4c78a1k6kn2f2.ris",
  "sources": [
    {
      "first_seen": "2026-10-30T12:23:33Z",
      "native_id": "2609.22226",
      "source": "arxiv",
      "url": "https://arxiv.org/abs/2609.22226"
    }
  ],
  "subjects": [
    {
      "field": "mathematics",
      "origin": "arxiv:stat.ML",
      "subfield": "statistics"
    }
  ],
  "title": "Swiss-Knife: A Framework for Reconfigurable Externalised Multi-Objective Alignment at Decode Time",
  "title_text": "Swiss-Knife: A Framework for Reconfigurable Externalised Multi-Objective Alignment at Decode Time",
  "type": "preprint",
  "updated_at": "2026-10-30T12:23:33Z",
  "url": "https://research.draconitic.com/paper/4c78a1k6kn2f2",
  "venue": "arXiv",
  "versions": []
}