{"paper":{"title":"MidSteer: Optimal Affine Framework for Steering Generative Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"MidSteer is a general affine framework for concept steering in generative models that relaxes optimality assumptions of prior LEACE-based methods to enable directed minimal-disturbance transformations.","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Andrew Stepanov, Gregory Slabaugh, Ismail Elezi, Jiankang Deng, Martin Benning, Tatiana Gaintseva, Ziquan Liu","submitted_at":"2026-04-17T19:23:33Z","abstract_excerpt":"Steering intermediate representations has emerged as a powerful strategy for controlling generative models, particularly in post-deployment alignment and safety settings. However, despite its empirical success, it currently lacks a comprehensive theoretical framework. In this paper, we bridge this gap by formalizing the theory of concept steering. First, we establish a link between steering and affine concept erasure, proving that the standard approach for removing unwanted behaviors is a special case of LEACE (a closed-form method for affine erasure). Next, we formulate a principled theoretic"},"claims":{"count":3,"items":[{"kind":"strongest_claim","text":"We introduce MidSteer (Minimal Disturbance concept Steering), a more general affine framework for concept manipulation that relaxes these assumptions and enables directed, minimal-disturbance transformations.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The assumptions under which LEACE-Switch provides an optimal affine solution hold for the specific concept manipulations considered; MidSteer relaxes them but still relies on affine transformations being sufficient for effective steering.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"MidSteer is a general affine framework for concept steering in generative models that relaxes optimality assumptions of prior LEACE-based methods to enable directed minimal-disturbance transformations.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"}],"snapshot_sha256":"480a624475eb24f82cd3214101bab33c8f0e72b958fac9fe4cfcef9d326f6033"},"source":{"id":"2605.05220","kind":"arxiv","version":2},"verdict":{"id":"26a76550-39ec-4423-85cf-0600d8b068dc","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T08:34:47.208240Z","strongest_claim":"We introduce MidSteer (Minimal Disturbance concept Steering), a more general affine framework for concept manipulation that relaxes these assumptions and enables directed, minimal-disturbance transformations.","one_line_summary":"MidSteer is a general affine framework for concept steering in generative models that relaxes optimality assumptions of prior LEACE-based methods to enable directed minimal-disturbance transformations.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The assumptions under which LEACE-Switch provides an optimal affine solution hold for the specific concept manipulations considered; MidSteer relaxes them but still relies on affine transformations being sufficient for effective steering.","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.05220/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}