AI & ML
Swiss-Knife: A Framework for Reconfigurable Externalised Multi-Objective Alignment at Decode Time
arXiv stat.ML0 views
arXiv:2609.22226v2 Announce Type: replace-cross 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 $R_i = \sum_{j \neq i} g((\mu_i - \mu_j)/s(\sigma_i,\sigma_j))$, 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.
Source:https://arxiv.org/abs/2609.22226PreviousBad Genius: Counterfactual-Guided Harness Evolution Beyond Task-Specific ShortcutsNextIQS-BO: In-Context Query Selection for Bayesian Optimisation
Related posts
Re-Testing LLM-Based C-to-Rust Translations Against Their Original C
0arXiv cs.SE
Reliability is a user-specific property of large language models
0arXiv cs.SE
Safe Personal RSI: Evaluating and Repairing Personal-Agent Updates
0arXiv cs.SE
Option-GRPO: Hierarchical RL Against AI Slop in Code Generation for Unseen SDKs
0arXiv cs.SE
Systematic Mapping Protocol. Generative AI-Augmented Requirements Engineering: A Systematic Mapping Study of Practices, Specifications, and Reuse in Software Development
0arXiv cs.SE
RootSink: Reproducible C/C++ Crashes with Tested Repair Sites
0arXiv cs.SE