Pith. sign in

REVIEW 5 cited by

What's the Magic Word? A Control Theory of LLM Prompting

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2310.04444 v4 pith:JOTOKILW submitted 2023-10-02 cs.CL cs.AIcs.LGcs.NE

What's the Magic Word? A Control Theory of LLM Prompting

classification cs.CL cs.AIcs.LGcs.NE
keywords leastlikelyllmsprompttokensanalysiscontrolcontrollability
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
read the original abstract

Prompt engineering is crucial for deploying LLMs but is poorly understood mathematically. We formalize LLM systems as a class of discrete stochastic dynamical systems to explore prompt engineering through the lens of control theory. We offer a mathematical analysis of the limitations on the controllability of self-attention as a function of the singular values of the parameter matrices. We present complementary empirical results on the controllability of a panel of LLMs, including Falcon-7b, Llama-7b, and Falcon-40b. Given initial state $\mathbf x_0$ from Wikitext and prompts of length $k \leq 10$ tokens, we find that the "correct" next token is reachable at least 97% of the time, and that the top 75 most likely next tokens are reachable at least 85% of the time. Intriguingly, short prompt sequences can dramatically alter the likelihood of specific outputs, even making the least likely tokens become the most likely ones. This control-theoretic analysis of LLMs demonstrates the significant and poorly understood role of input sequences in steering output probabilities, offering a foundational perspective for enhancing language model system capabilities.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 5 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Activation Steering of Video Generation Models via Reduced-Order Linear Optimal Control

    cs.LG 2026-06 unverdicted novelty 7.0

    LA-LQR applies latent-space linear-quadratic regulator control to steer text-to-video model activations toward desired features while penalizing excessive changes.

  2. Local Linearity of LLMs Enables Activation Steering via Model-Based Linear Optimal Control

    cs.LG 2026-04 conditional novelty 7.0

    Local linearity of LLM layers enables LQR-based closed-loop activation steering with theoretical tracking guarantees.

  3. FDM-Bench: A Comprehensive Benchmark for Evaluating Large Language Models in Additive Manufacturing Tasks

    cs.LG 2024-12 conditional novelty 7.0

    FDM-Bench is a new benchmark dataset for evaluating LLMs on FDM tasks including user queries and G-code anomaly detection, with expert-assessed results showing closed-source models outperforming on anomaly detection a...

  4. Steering Robustness into World Action Models via Mechanistic Interpretability and Optimal Control

    cs.RO 2026-07 conditional novelty 6.0

    Contrastive activation directions and a reduced-order LQR (WA-LQR) steer world-action models to recover robustness under camera, gripper, and noise shifts whenever the models' activations are linearly separable for th...

  5. When control meets large language models: From words to dynamics

    eess.SY 2026-02 unverdicted novelty 3.0

    The paper proposes a bidirectional continuum between LLMs and control systems, covering LLM-assisted controller design, control-based LLM steering, and state-space modeling of LLMs.