arXiv subjects

Jesper Ferkinghoff-Borg

Publications and source records attributed to Jesper Ferkinghoff-Borg.

13 recordsLinked to original sources

arXiv2026

Measuring Black-Box Confidence via Reasoning Trajectories: Geometry, Coverage, and Verbalization

Reliable confidence estimation enables safe deployment of chain-of-thought (CoT) reasoning through text-only APIs. Yet the dominant black-box baseline, self-consistency over K samples, is linearly expensive and ignores the geometry of the trace. We propose a black-box trajectory-confidence score: we embed a CoT as a sliding-window trajectory and measure its convergence to external answer anchors with a one-parameter softmax. The method needs no logits, hidden states, or supervised calibrators. Across six (benchmark, reasoner) settings on MedQA-USMLE, GPQA Diamond, and MMLU-Pro with Gemini 3.1 Pro and Claude Sonnet 4.6, fusing this score with coverage and verbalized-confidence channels at K=4 yields Pareto improvements over self-consistency at K=8 in 6/6 settings (median AUC 0.78 vs 0.71, deltaAUC=+0.075). A fixed-pick control (+0.060) and E5 cross-embedder replication rule out answer switching and single-vendor artifacts. Geometry peaks in the penultimate window across benchmarks and reasoners, and inverts at the terminal window on GPQA Diamond. Three unscaffolded regimes separate black-box confidence into a judge-mediated Coverage prior (C), within-trace Geometry (G), and a conditional Verbalization channel (V). Across 18 benchmark x reasoner x proposer settings, C and G provide independent signal in 18/18 and 16/18, while V contributes residual signal in 6/18. Swapping the judge from GPT-5-mini to Claude Sonnet 4.6 leaves G-only AUC unchanged (|delta|<=0.013) and shifts C-only AUC by at most +/-0.02 (kappa=0.82). Fusion beats the best single channel in 17/18 settings (median AUC 0.78, max 0.92).

cs.AI↗

arXiv2026

MechPert: Mechanistic Consensus as an Inductive Bias for Unseen Perturbation Prediction

Predicting transcriptional responses to unseen genetic perturbations is essential for understanding gene regulation and prioritizing large-scale perturbation experiments. Existing approaches either rely on static, potentially incomplete knowledge graphs, or prompt language models for functionally similar genes, retrieving associations shaped by symmetric co-occurrence in scientific text rather than directed regulatory logic. We introduce MechPert, a lightweight framework that encourages LLM agents to generate directed regulatory hypotheses rather than relying solely on functional similarity. Multiple agents independently propose candidate regulators with associated confidence scores; these are aggregated through a consensus mechanism that filters spurious associations, producing weighted neighborhoods for downstream prediction. We evaluate MechPert on Perturb-seq benchmarks across four human cell lines. For perturbation prediction in low-data regimes ($N=50$ observed perturbations), MechPert improves Pearson correlation by up to 10.5\% over similarity-based baselines. For experimental design, MechPert-selected anchor genes outperform standard network centrality heuristics by up to 46\% in well-characterized cell lines.

cs.LG↗

arXiv2025

Towards Agents That Know When They Don't Know: Uncertainty as a Control Signal for Structured Reasoning

Large language model (LLM) agents are increasingly deployed in structured biomedical data environments, yet they often produce fluent but overconfident outputs when reasoning over complex multi-table data. We introduce an uncertainty-aware agent for query-conditioned multi-table summarization that leverages two complementary signals: (i) retrieval uncertainty--entropy over multiple table-selection rollouts--and (ii) summary uncertainty--combining self-consistency and perplexity. Summary uncertainty is incorporated into reinforcement learning (RL) with Group Relative Policy Optimization (GRPO), while both retrieval and summary uncertainty guide inference-time filtering and support the construction of higher-quality synthetic datasets. On multi-omics benchmarks, our approach improves factuality and calibration, nearly tripling correct and useful claims per summary (3.0$\rightarrow$8.4 internal; 3.6$\rightarrow$9.9 cancer multi-omics) and substantially improving downstream survival prediction (C-index 0.32$\rightarrow$0.63). These results demonstrate that uncertainty can serve as a control signal--enabling agents to abstain, communicate confidence, and become more reliable tools for complex structured-data environments.

cs.AI↗

arXiv2025

Zero-shot protein stability prediction by inverse folding models: a free energy interpretation

Inverse folding models have proven to be highly effective zero-shot predictors of protein stability. Despite this success, the link between the amino acid preferences of an inverse folding model and the free-energy considerations underlying thermodynamic stability remains incompletely understood. A better understanding would be of interest not only from a theoretical perspective, but also potentially provide the basis for stronger zero-shot stability prediction. In this paper, we take steps to clarify the free-energy foundations of inverse folding models. Our derivation reveals the standard practice of likelihood ratios as a simplistic approximation and suggests several paths towards better estimates of the relative stability. We empirically assess these approaches and demonstrate that considerable gains in zero-shot performance can be achieved with fairly simple means.

cs.LG↗

arXiv2024

Batched Energy-Entropy acquisition for Bayesian Optimization

Bayesian optimization (BO) is an attractive machine learning framework for performing sample-efficient global optimization of black-box functions. The optimization process is guided by an acquisition function that selects points to acquire in each round of BO. In batched BO, when multiple points are acquired in parallel, commonly used acquisition functions are often high-dimensional and intractable, leading to the use of sampling-based alternatives. We propose a statistical physics inspired acquisition function for BO with Gaussian processes that can natively handle batches. Batched Energy-Entropy acquisition for BO (BEEBO) enables tight control of the explore-exploit trade-off of the optimization process and generalizes to heteroskedastic black-box problems. We demonstrate the applicability of BEEBO on a range of problems, showing competitive performance to existing methods.

cs.LG↗

arXiv2022

Probabilistic thermal stability prediction through sparsity promoting transformer representation

Pre-trained protein language models have demonstrated significant applicability in different protein engineering task. A general usage of these pre-trained transformer models latent representation is to use a mean pool across residue positions to reduce the feature dimensions to further downstream tasks such as predicting bio-physics properties or other functional behaviours. In this paper we provide a two-fold contribution to machine learning (ML) driven drug design. Firstly, we demonstrate the power of sparsity by promoting penalization of pre-trained transformer models to secure more robust and accurate melting temperature (Tm) prediction of single-chain variable fragments with a mean absolute error of 0.23C. Secondly, we demonstrate the power of framing our prediction problem in a probabilistic framework. Specifically, we advocate for the need of adopting probabilistic frameworks especially in the context of ML driven drug design.

stat.ML↗

arXiv2010

Potentials of Mean Force for Protein Structure Prediction Vindicated, Formalized and Generalized

Understanding protein structure is of crucial importance in science, medicine and biotechnology. For about two decades, knowledge based potentials based on pairwise distances -- so-called "potentials of mean force" (PMFs) -- have been center stage in the prediction and design of protein structure and the simulation of protein folding. However, the validity, scope and limitations of these potentials are still vigorously debated and disputed, and the optimal choice of the reference state -- a necessary component of these potentials -- is an unsolved problem. PMFs are loosely justified by analogy to the reversible work theorem in statistical physics, or by a statistical argument based on a likelihood function. Both justifications are insightful but leave many questions unanswered. Here, we show for the first time that PMFs can be seen as approximations to quantities that do have a rigorous probabilistic justification: they naturally arise when probability distributions over different features of proteins need to be combined. We call these quantities reference ratio distributions deriving from the application of the reference ratio method. This new view is not only of theoretical relevance, but leads to many insights that are of direct practical use: the reference state is uniquely defined and does not require external physical insights; the approach can be generalized beyond pairwise distances to arbitrary features of protein structure; and it becomes clear for which purposes the use of these quantities is justified. We illustrate these insights with two applications, involving the radius of gyration and hydrogen bonding. In the latter case, we also show how the reference ratio method can be iteratively applied to sculpt an energy funnel. Our results considerably increase the understanding and scope of energy functions derived from known biomolecular structures.

q-bio.BM↗

arXiv2009

Stop-and-go kinetics in amyloid fibrillation

Many human diseases are associated with protein aggregation and fibrillation. We present experiments on in vitro glucagon fibrillation using total internal reflection fluorescence microscopy, providing real-time measurements of single-fibril growth. We find that amyloid fibrils grow in an intermittent fashion, with periods of growth followed by long pauses. The observed exponential distributions of stop and growth times support a Markovian model, in which fibrils shift between the two states with specific rates. Remarkably, the probability of being in the growing (stopping) state is very close to 1/4 (3/4) in all experiments, even if the rates vary considerably. This finding suggests the presence of 4 independent conformations of the fibril tip; we discuss this possibility in terms of the existing structural knowledge.

physics.bio-ph↗

arXiv2005

Diffusion, Fragmentation and Coagulation Processes: Analytical and Numerical Results

We formulate dynamical rate equations for physical processes driven by a combination of diffusive growth, size fragmentation and fragment coagulation. Initially, we consider processes where coagulation is absent. In this case we solve the rate equation exactly leading to size distributions of Bessel type which fall off as $\exp(-x^{3/2})$ for large $x$-values. Moreover, we provide explicit formulas for the expansion coefficients in terms of Airy functions. Introducing the coagulation term, the full non-linear model is mapped exactly onto a Riccati equation that enables us to derive various asymptotic solutions for the distribution function. In particular, we find a standard exponential decay, $\exp(-x)$, for large $x$, and observe a crossover from the Bessel function for intermediate values of $x$. These findings are checked by numerical simulations and we find perfect agreement between the theoretical predictions and numerical results.

cond-mat.stat-mech↗

arXiv2005

Scale Free Cluster Distributions from Conserving Merging-Fragmentation Processes

We propose a dynamical scheme for the combined processes of fragmentation and merging as a model system for cluster dynamics in nature and society displaying scale invariant properties. The clusters merge and fragment with rates proportional to their sizes, conserving the total mass. The total number of clusters grows continuously but the full time-dependent distribution can be rescaled over at least 15 decades onto a universal curve which we derive analytically. This curve includes a scale free solution with a scaling exponent of -3/2 for the cluster sizes.

cond-mat.stat-mech↗

arXiv2004

Dynamics of Crystal Formation in the Greenland NorthGRIP Ice Core

The North Greenland Ice Core Project (NorthGRIP) provides paleoclimatic information back to at about 120 kyr before present (Dahl-Jensen and others, 2002). Each year, precipitation on the ice sheet covers it with a new layer of snow, which gradually transforms into ice crystals as the layer sinks into the ice sheet. The size distribution of ice crystals has been measured at selected depths in the upper 880 m of the NorthGRIP ice core (Svensson and others, 2003b), which covers a time span of 5300 years. The distributions change with time toward a universal curve, indicating a common underlying physical process in the formation of crystals. We identify this process as an interplay between fragmentation of the crystals and diffusion of their grain boundaries. The process is described by a two-parameter differential equation to which we obtain the exact solution. The solution is in excellent agreement with the experimentally observed distributions.

physics.geo-ph↗

arXiv2003

Competition between Diffusion and Fragmentation: An Important Evolutionary Process of Nature

We investigate systems of nature where the common physical processes diffusion and fragmentation compete. We derive a rate equation for the size distribution of fragments. The equation leads to a third order differential equation which we solve exactly in terms of Bessel functions. The stationary state is a universal Bessel distribution described by one parameter, which fits perfectly experimental data from two very different system of nature, namely, the distribution of ice crystal sizes from the Greenland ice sheet and the length distribution of alpha-helices in proteins.

cond-mat.stat-mech↗

arXiv2003

Compact phases of polymers with hydrogen bonding

We propose an off-lattice model for a self-avoiding homopolymer chain with two different competing attractive interactions, mimicking the hydrophobic effect and the hydrogen bond formation respectively. By means of Monte Carlo simulations, we are able to trace out the complete phase diagram for different values of the relative strength of the two competing interactions. For strong enough hydrogen bonding, the ground state is a helical conformation, whereas with decreasing hydrogen bonding strength, helices get eventually destabilized at low temperature in favor of more compact conformations resembling $β$-sheets appearing in native structures of proteins. For weaker hydrogen bonding helices are not thermodynamically relevant anymore.

cond-mat.stat-mech↗