Knob: A Physics-Inspired Gating Interface for Interpretable and Controllable Neural Dynamics

Siyu Jiang; Sanshuai Cui; Hui Zeng

cs.AIFeb 26, 2026

Knob: A Physics-Inspired Gating Interface for Interpretable and Controllable Neural Dynamics

Authors: Siyu Jiang, Sanshuai Cui, Hui Zeng

Organizations: City University of Macau, Macau, China · Southwest University of Science and Technology, Mianyang, China

Abstract

Existing neural network calibration methods often treat calibration as a static, post-hoc optimization task. However, this neglects the dynamic and temporal nature of real-world inference. Moreover, existing methods do not provide an intuitive interface enabling human operators to dynamically adjust model behavior under shifting conditions. In this work, we propose Knob, a framework that connects deep learning with classical control theory by mapping neural gating dynamics to a second-order mechanical system. By establishing correspondences between physical parameters -- damping ratio ( $ζ$ ) and natural frequency ( $ω_n$ ) -- and neural gating, we create a tunable "safety valve". The core mechanism employs a logit-level convex fusion, functioning as an input-adaptive temperature scaling. It tends to reduce model confidence particularly when model branches produce conflicting predictions. Furthermore, by imposing second-order dynamics (Knob-ODE), we enable a dual-mode inference: standard i.i.d. processing for static tasks, and state-preserving processing for continuous streams. Our framework allows operators to tune "stability" and "sensitivity" through familiar physical analogues. This paper presents an exploratory architectural interface; we focus on demonstrating the concept and validating its control-theoretic properties rather than claiming state-of-the-art calibration performance. Experiments on CIFAR-10-C validate the calibration mechanism and demonstrate that, in Continuous Mode, the gate responses are consistent with standard second-order control signatures (step settling and low-pass attenuation), paving the way for predictable human-in-the-loop tuning.

Figures & tables

Figure 1: The Knob framework as a physics-inspired interface. Left (Dual-Stream Backbone): A shared encoder feeds two lightweight projection heads, producing a robust “Static” branch and a more sensitive “Dynamic” branch, yielding complementary logit vectors $\bm{z}_{\mathrm{static}}$ and $\bm{z}_{\mathrm{dyn}}$ . Center (Physics Engine): The gating mechanism is modeled as a mass-spring-damper system (normalized to unit mass). A network-predicted reference input $u^{*}(x)$ drives the system; its response is governed by two interpretable control parameters— Natural Frequency ( $\omega_{n}$ ), controlling sensitivity/bandwidth, and Damping Ratio ( $\zeta$ ), controlling stability/conservativeness—which act as tunable “knobs” for the operator. Right (Convex Fusion): The gate value $g(x)\in[0,1]$ , determined by the mass displacement, performs a convex combination of two logit branches, naturally limiting model overconfidence.

Symbol/Term	Definition and Intuition
Methodology Symbols
$z_{\text{static}},z_{\text{dyn}}$	Logit vectors from the static and dynamic branches, respectively.
$g(x)$	A sample-wise scalar gate value in $[0,1]$ that controls the convex fusion.
$z_{\text{fuse}}$	The final logit vector after convex fusion.
$T_{\mathrm{eff}}(x)$	The effective temperature; a measure of induced smoothing ( $T_{\mathrm{eff}}\geq 1$ ).
$\zeta,\omega_{n}$	Damping ratio and natural frequency; control overshoot and convergence speed.

Table 1: Terminology and Metric Definitions.

Method	Avg-C (%) $\uparrow$	Err-C (%) $\downarrow$	ECE deb $\downarrow$	Latency (ms)
Static	71.25	28.75	0.1553	3.48
Attention	70.45	29.55	0.1648	3.51
Knob-IA	72.52	27.48	0.1617	3.49
ODE-Lite	71.62	28.38	0.1488	3.64
Knob-ODE	71.49	28.51	0.1829	4.41

Table 2: Main results on CIFAR-10-C (Reset Mode). Calibration and efficiency under identical budgets (3-run averages). Lower is better for Err-C and ECE deb . Avg-C = mean accuracy averaged over severities and corruptions.

Panel of Figure 2 — Figure 2: Cost–calibration Pareto frontiers. Left: ECE deb vs. GFLOPs. Right: ECE deb vs. latency. ODE-Lite lies near the frontier in both views, indicating a favorable calibration–cost trade-off.

Method	Params (M)	GFLOPs	Latency (ms)
Static	11.70	0.5584	3.48
Attention	11.90	0.5586	3.51
Knob-IA	11.77	0.5585	3.49
ODE-Lite	11.77	0.5585	3.64
Knob-ODE	11.77	0.5585	4.41

Table 3: Efficiency comparison. The gate controller introduces only

\mathcal{O}(D)

incremental overhead, and overall computational costs are essentially identical across methods.

Panel of Figure 3 — Figure 3: E-1: Static equivalence evidence (Knob-IA). (a) CSR for the overall, disagreement, and agreement subsets stays below $1$ and decreases with corruption severity. (b) $\widehat{T}(x)\!\geq\!1$ and increases with severity (Spearman’s $\rho=0.6$ ). (c) Reliability diagram comparing disagreement and agreement subsets. (d) Per-bin ECE by confidence bins.

Figure 4: E-2: Learning-level probe—gate “taking sides.” Treating $g(x)$ as a “dynamic-preferred” score, we plot AUC for Oracle Advantage (OA, solid lines) and Gradient-Consistent Advantage (GCA, dashed lines) across different sample subsets (overall, top-1 disagreement, margin sign). The polarization index $\Pi=\mathbb{E}[\,|g-0.5|\,]$ is shown on the right axis (gray triangles). AUC steadily increases during training, with overall $\mathrm{AUC}_{\mathrm{OA}}\approx 0.79$ and $\mathrm{AUC}_{\mathrm{GCA}}\approx 0.74$ by epoch 50, while $\Pi$ remains low ( $\approx 0.10$ ), indicating the gate learns to discriminate without saturating.

Figure 5: E-3: Dynamics-level probe (Continuous Mode). (a) Step Response: Under a shot-noise severity schedule ( $1\to 5\to 1$ ), the gate $g(t)$ (blue) exhibits a smooth, damped transition rather than an instantaneous jump. Shaded bands indicate the inter-quartile range (IQR) over stochastic corruption realizations. (b) Empirical Bode Plot: The magnitude response relative to DC (0 dB) shows a downward trend as frequency increases, consistent with soft low-pass filtering behavior. Error bars represent standard error of the mean (SEM).

Model Calibration

Model calibration aligns predicted confidence or model outputs with observed outcomes and measurements to support more reliable decisions.

See all 214 papers

Calibration in One Go: Jointly Learning Accurate and Calibrated Neural Networks
Oct 9, 2026Nektarios Kalampalikis, Pablo M. Olmos, Isabel ValeraModel Calibration Post-Hoc Calibration
Dept. of Computer Science Saarland University, Germany
Learning Inference Concurrency in DynamicGate MLP Structural and Mathematical Justification
Apr 15, 2026Yongil ChoiDynamic Neural Networks Online Learning
Sorynorydotcom Co., Ltd./AI Open Research Lab
Too Sharp, Too Sure: When Calibration Follows Curvature
Apr 22, 2026Alessandro Morosini, Matea Gjika, Tomaso Poggio +1Loss Landscape Sharpness Margin-Based Learning
Massachusetts Institute of Technology, Cambridge, MA, USA.

Oct 9, 2026J/K move · Enter open · S save

Calibration in One Go: Jointly Learning Accurate and Calibrated Neural Networks

Nektarios Kalampalikis, Pablo M. Olmos, Isabel Valera

Dept. of Computer Science Saarland University, Germany

Model Calibration Post-Hoc Calibration

Open paper AskPDF arXiv

Neural networks (NNs), while highly flexible and versatile, are often poorly calibrated. Post-hoc calibration methods adjust the outputs of an already trained model using a separate calibration dataset and additional calibration parameters. This process adds overhead that requires separate optimization of the calibrator and subsequent model selection. In this paper, we show that standard (post-hoc) calibrators can be trained end-to-end with the backbone via a bilevel Stackelberg formulation. This formulation eliminates the need for a separate calibration set or a dedicated calibration loss by jointly optimizing accuracy and calibration using cross-entropy, introducing only one additional hyperparameter: the calibrator learning rate. Although our framework supports calibrators ranging from temperature and matrix scaling to NN layers, our results show that using the final classification layer itself as the calibrator yields accurate, well-calibrated models with no additional parameters. Our approach outperforms SOTA calibrators with minimal added complexity, while also avoiding the multiplicative overhead of separate calibration components in settings such as multi-task learning.

Abstract

Model Calibration

Explore similar work

Calibration in One Go: Jointly Learning Accurate and Calibrated Neural Networks

Learning Inference Concurrency in DynamicGate MLP Structural and Mathematical Justification

Too Sharp, Too Sure: When Calibration Follows Curvature