{"paper":{"title":"Adaptive Activation Steering for Efficient LLM Reasoning via Closed-Loop PID Control","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"A PID controller dynamically adjusts activation steering to cut redundant chain-of-thought steps in large language models.","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Aryasomayajula Ram Bharadwaj","submitted_at":"2025-06-23T16:47:19Z","abstract_excerpt":"Reasoning LLMs trained with long chain-of-thought often overthink: they spend tokens on redundant reflection and transitions that inflate cost without improving accuracy. Static activation steering (e.g.\\ SEAL) suppresses such content with a fixed vector, but applies the same strength regardless of how redundant the current chunk actually is. We describe PID-steering, a training-free, decoding-time method that modulates the steering strength with a PID controller driven by a lightweight chunk-level redundancy classifier. On a subset of GSM8K with DeepSeek-R1-Distill-Qwen-1.5B, the method impro"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Experimental evaluation on GSM8K demonstrates that STUPID achieves a 6% improvement in accuracy while reducing token usage by 32%, outperforming static steering baselines.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The chunk-level classifier reliably detects redundant reasoning patterns in real time and that its redundancy probability can be used as a stable error signal for the PID controller without introducing new failure modes.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"STU-PID combines a chunk-level redundancy classifier with a PID controller to dynamically modulate activation steering in LLMs, yielding 6% higher accuracy and 32% fewer tokens on GSM8K compared to static baselines.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A PID controller dynamically adjusts activation steering to cut redundant chain-of-thought steps in large language models.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"adf579ab6265be845723eebbf1e5633bb3c17822e9240c62dbd92d57011e0312"},"source":{"id":"2506.18831","kind":"arxiv","version":3},"verdict":{"id":"15efd949-bac2-45fa-a367-4828f769767a","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-19T07:52:07.170603Z","strongest_claim":"Experimental evaluation on GSM8K demonstrates that STUPID achieves a 6% improvement in accuracy while reducing token usage by 32%, outperforming static steering baselines.","one_line_summary":"STU-PID combines a chunk-level redundancy classifier with a PID controller to dynamically modulate activation steering in LLMs, yielding 6% higher accuracy and 32% fewer tokens on GSM8K compared to static baselines.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The chunk-level classifier reliably detects redundant reasoning patterns in real time and that its redundancy probability can be used as a stable error signal for the PID controller without introducing new failure modes.","pith_extraction_headline":"A PID controller dynamically adjusts activation steering to cut redundant chain-of-thought steps in large language models."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2506.18831/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"}