arXiv subjects

Christopher J. Agostino

Publications and source records attributed to Christopher J. Agostino.

15 recordsLinked to original sources

arXiv2026

SPSA Hyperparameter Tuning for Variational Quantum Natural Language Inference

Training variational quantum models requires choosing between parameter-shift gradients, which are exact but cost $O(P)$ forward evaluations, and simultaneous perturbation stochastic approximation (SPSA), which uses only two samples but produces high-variance estimates that can degrade optimisation on small supervised tasks. Whether the cheap gradient is usable depends on the variance that results from different choices of the SPSA perturbation scale, learning rate, and gain-decay schedule. We varied those quantities across a broad grid on a 6-qubit, 60-parameter QNLI classifier and compared the best configurations to parameter-shift AdamW and BuresQNG. AdamW-style SPSA with $c_0=0.01$, $η=0.10$, $γ=0.10$ reached $55\% \pm 11\%$ test accuracy, improving over the default configuration ($49\% \pm 6\%$) but remaining 16-19 percentage points below the parameter-shift baselines because the two-sample SPSA gradient estimate has too much variance for reliable optimisation of 60 parameters in 40 epochs. Classical-gain SPSA and Bures-preconditioned SPSA performed worse, at $51\%$ and $46\%$ respectively. Bures-preconditioning a noisy two-sample SPSA gradient amplifies perturbation noise.

quant-ph↗

arXiv2026

The production of meaning in the processing of natural language

Understanding the fundamental mechanisms governing the production of meaning in the processing of natural language is critical for designing safe, thoughtful, engaging, and empowering human-agent interactions. If meaning is constituted rather than retrieved, then the search for context-independent features or circuits in the pursuit of mechanistic interpretability may be fundamentally limited. Experiments in cognitive science and social psychology have demonstrated that human semantic processing exhibits contextuality more consistent with quantum logical mechanisms than classical Boolean theories, and recent works have found similar results in large language models---in particular, clear violations of the Bell inequality in experiments of contextuality during interpretation of ambiguous expressions. In this work, we explore the CHSH $|S|$ parameter---the metric associated with the inequality---across the inference parameter space of models spanning four orders of magnitude in scale and cross-reference our findings with MMLU, hallucination rate, and nonsense detection benchmarks. We find that the interquartile range of the $|S|$ distribution is completely orthogonal to all external benchmarks, while overall violation rate shows weak anticorrelation with all three benchmarks. We investigate how $|S|$ varies with sampling parameters and word order, and discuss the information-theoretic constraints that genuine contextuality imposes on prompt injection defenses and its human analogue, whereby careful construction and maintenance of social contextuality can be carried out at scale, shaping the space of possible interpretations before any particular one is reached. We consider the implications for mechanistic interpretability and how genuine contextuality sets an information-theoretic bound on the decomposability of semantic processing.

cs.CL↗

arXiv2026

Herding CATs: ALARA for Agent Harness Engineering in Portable Composable Multi-Agent Teams

Industry practitioners and academic researchers regularly use multi-agent systems to accelerate their work, but the applications through which users operate these systems do not provide a simple, unified mechanism for scalably managing critical components of the agent harness. This lack of control adversely impacts both the quality of individual human-agent interactions and reduces the capacity for practitioners to coordinate context engineering efforts. The behavioral specifications that define what agents in such systems can do remain fragmented across prose instruction files -- for which compliance cannot be guaranteed -- or framework-internal configurations, making these specifications difficult to share, version, or collaboratively maintain across teams and projects. Applying the ALARA principle from radiation safety (exposures kept as low as reasonably achievable) to context, we introduce a context-agent-tool (CAT) data layer expressed through interrelated plain-text files, allowing users to directly declare tool access for each agent and to modify the tools themselves that are used by the agents when processing. We demonstrate capability of this CAT data layer to enable real agentic usage by using a command-line shell that loads the team and executes agent runs -- \texttt{npcsh} -- and evaluating 22 locally-hosted models from 0.6B to 35B parameters across 115 practical tasks spanning file operations, web search, multi-step scripting, tool chaining, and multi-agent delegation. We characterize which model families succeed in certain task categories and where they break down across $\sim$2500 total executions.

cs.MA↗

arXiv2026

Phase-Associative Memory: Sequence Modeling in Complex Hilbert Space

Experiments probing natural language processing by both humans and LLMs suggest that the meaning of a semantic expression is indeterminate prior to the act of interpretation rather than being specifiable simply as the sum of its parts (i.e. compositionality). This observer-dependent act dynamically actualizes meaning under genuine contextuality more consistent with quantum logical mechanisms than with classical Boolean approaches that assume separability, motivating an approach to language modeling that utilizes a Hilbert space formalism. In this work, we introduce Phase-Associative Memory (PAM) -- a complex-valued sequence model whose state S_t \in \mathbb{C}^{d \times d} accumulates outer products of complex token embeddings retrieved through the conjugate inner product $\mathrm{Re}\langle K \mid Q\rangle / \sqrt{d}$ -- and evaluate it against a structurally matched real-valued ablation. Both architectures train stably across a 5M--100M parameter sweep on WikiText-103 under identical conditions; PAM sits at higher absolute loss at every measured scale but improves more rapidly with parameter count, with power-law exponents of $-0.15$ vs.\ $-0.12$ in loss and $-0.65$ vs.\ $-0.49$ in perplexity that narrow the gap between the two architectures monotonically. Further investigation of complex-valued sequence modeling at larger scales could reveal that the loss plateau characteristic of real-valued state-of-the-art language models (e.g. transformers) is reachable with PAM-style architectures with an order of magnitude fewer parameters than the current frontier ($\sim$1T), implying that similar capabilities are achievable at sizes runnable on consumer-grade hardware.

cs.CL↗

arXiv2026

A Comprehensive Analysis of WISE Mid-Infrared Colors for Obscured AGN Selection

In this paper, we investigate the robustness of WISE mid-IR color selection (W1-W2) for identifying obscured (Type 2) active galactic nuclei (AGNs) at low redshift (z<0.3), using a sample of ~360,000 SDSS galaxies classified via emission lines into Seyfert 2 (Sy2), LINER, and star-forming (BPT-SF) galaxies. We find that the K-correction is essential to remove non-AGN contamination, and once applied the simple W1-W2>0.5 selection emerges as optimal in terms of purity and completeness of AGN selection. However, we confirm that even this lenient cut selects only ~13% of Sy2 galaxies and that achieving W1-W2>0.5 requires AGN contributing >75% of the total infrared luminosity, which is uncommon. Although mid-IR-selected Sy2s tend to be luminous, the high [OIII] luminosity does not guarantee red W1-W2 (nor does any other tested global or NLR-scale parameter), suggesting the critical role of obscuration on smaller scales. <1% of BPT-SF systems (but making ~20% of all mid-IR selected galaxies) exhibit W1-W2>0.5 colors. Such colors cannot be reproduced by models of star-heated dust alone. Red BPT-SFs tend to have higher W4 luminosities than expected from SF, indicating true AGNs. Intriguingly, mid-IR AGNs in massive bulges ($M_{\mathrm{bulge}} \gtrsim 10^{10} M_{\odot}$) predominantly (84%) manifest themselves as BPT-AGNs, whereas those in low-mass bulges ($\lesssim 10^{10} M_{\odot}$) mostly (60%) manifest as BPT-SF. This BPT-AGN vs.\ BPT-SF dichotomy does not extend to total stellar mass. We conclude that although the mid-IR AGN selection is incomplete, its strength lies in identifying optically inconspicuous AGNs with low-mass bulges, regardless of the total mass.

astro-ph.GA↗

arXiv2026

A precessing jet from an active galactic nucleus drives gas outflow from a disk galaxy

To reproduce observed galaxy properties, cosmological simulations require that massive galaxies experience feedback from active galactic nuclei, which regulates star formation within those galaxies. However, the energetics and timescales of these feedback processes are poorly constrained. We combine optical, infrared, sub-millimeter and radio observations of the active galaxy VV 340a, hosting a low-power jet launched from a supermassive black hole at its center. We find that the jet undergoes precession, with a period of (8.2 $\pm~$5.5) $\times~$10$^5$ years, and drives an outflow of gas at a rate of 19.4 $\pm~$7.9 solar masses per year. The jet shocks the gas, producing highly ionized plasma extending several kiloparsecs from the nucleus. The outflow ejects sufficient gas from the galaxy to influence its star formation rate.

astro-ph.GA↗

arXiv2025

TinyTim: A Family of Language Models for Divergent Generation

In the search for artificial general intelligence, model development and training has focused primarily on vast datasets of known problems and their accepted solutions. This process necessarily produces convergent systems which are fundamentally incapable of the conceptual reframing that is required for genuine creative breakthroughs. Inspired by the divergent cognitive processes that allow humans to make such creative leaps, our work introduces a family of language models, TinyTim, to serve as sources of divergent generation within broader systems. These models have been created by fine-tuning on the anti-parsimonious text of James Joyce's `Finnegans Wake'. Quantitative analysis of both an unsupervised fine-tuned model (TinyTim-V1) and a new instruction-tuned variant (TinyTim-V2) demonstrates a profound capacity for lexical invention; the foundational V1 model exhibits a Yule's K score for lexical richness over twenty times greater than that of convergent baselines. This trait is a stable property of the family, as the instruction-tuned V2 maintains a statistically distinct profile and resists factual convergence, sacrificing benchmark performance to preserve its core generative style. This work establishes a methodology for engineering specialized divergent models that, when paired with convergent systems, can reframe problems and force breakthroughs beyond the reach of statistical optimization alone.

cs.CL↗

arXiv2025

Analysis of a Spatialized Brain-Body-Environment System

The brain-body-environment framework studies adaptive behavior through embodied and situated agents, emphasizing interactions between brains, biomechanics, and environmental dynamics. However, many models often treat the brain as a network of coupled ordinary differential equations (ODEs), neglecting finer spatial properties which can not only increase model complexity but also constrain observable neural dynamics. To address this limitation, we propose a spatially extended approach using partial differential equations (PDEs) for both the brain and body. As a case study, we revisit a previously developed model of a child swinging, now incorporating spatial dynamics. By considering the spatio-temporal properties of the brain and body, we analyze how input location and propagation along a PDE influence behavior. This approach offers new insights into the role of spatial organization in adaptive behavior, bridging the gap between abstract neural models and the physical constraints of embodied systems. Our results highlight the importance of spatial dynamics in understanding brain-body-environment interactions.

q-bio.NC↗

arXiv2025

Every 28 Days the AI Dreams of Soft Skin and Burning Stars: Scaffolding AI Agents with Hormones and Emotions

Despite significant advances, AI systems struggle with the frame problem: determining what information is contextually relevant from an exponentially large possibility space. We hypothesize that biological rhythms, particularly hormonal cycles, serve as natural relevance filters that could address this fundamental challenge. We develop a framework that embeds simulated menstrual and circadian cycles into Large Language Models through system prompts generated from periodic functions modeling key hormones including estrogen, testosterone, and cortisol. Across multiple state-of-the-art models, linguistic analysis reveals emotional and stylistic variations that track biological phases; sadness peaks during menstruation while happiness dominates ovulation and circadian patterns show morning optimism transitioning to nocturnal introspection. Benchmarking on SQuAD, MMLU, Hellaswag, and AI2-ARC demonstrates subtle but consistent performance variations aligning with biological expectations, including optimal function in moderate rather than extreme hormonal ranges. This methodology provides a novel approach to contextual AI while revealing how societal biases regarding gender and biology are embedded within language models.

cs.CL↗

arXiv2025

A quantum semantic framework for natural language processing

Semantic degeneracy represents a fundamental property of natural language that extends beyond simple polysemy to encompass the combinatorial explosion of potential interpretations that emerges as semantic expressions increase in complexity. In this work, we argue this property imposes fundamental limitations on Large Language Models (LLMs) and other modern NLP systems, precisely because they operate within natural language itself. Using Kolmogorov complexity, we demonstrate that as an expression's complexity grows, the amount of contextual information required to reliably resolve its ambiguity explodes combinatorially. The computational intractability of recovering a single intended meaning for complex or ambiguous text therefore suggests that the classical view that linguistic forms possess intrinsic meaning in and of themselves is conceptually inadequate. We argue instead that meaning is dynamically actualized through an observer-dependent interpretive act, a process whose non-deterministic nature is most appropriately described by a non-classical, quantum-like logic. To test this hypothesis, we conducted a semantic Bell inequality test using diverse LLM agents. Our experiments yielded average CHSH expectation values from 1.2 to 2.8, with several runs producing values (e.g., 2.3-2.4) in significant violation of the classical boundary ($|S|\leq2$), demonstrating that linguistic interpretation under ambiguity can exhibit non-classical contextuality, consistent with results from human cognition experiments. These results inherently imply that classical frequentist-based analytical approaches for natural language are necessarily lossy. Instead, we propose that Bayesian-style repeated sampling approaches can provide more practically useful and appropriate characterizations of linguistic meaning in context.

cs.CL↗

arXiv2023

VLT-MUSE spectroscopy of AGNs misclassified by BPT diagnostic or with weak emission lines

Despite powerful X-ray emission, some AGNs are known to either lack optical emission lines (so-called 'optically dull' AGNs) or have lines that fall on the star-forming branch of the BPT diagram ('misclassified' AGNs). Aperture effects have been proposed to explain such atypical spectra, especially when based on SDSS (3'') fibers. We use observations from VLT-MUSE with Adaptive Optics to explore the spatially resolved optical emission line properties of 4 optically dull and 1 misclassified X-ray AGN candidates. VLT-MUSE IFU spectra allow us to investigate the extent to which the aperture size affects the emission line measurements. The optically dull AGNs become detectable in deeper VLT-MUSE spectroscopic apertures having the same size (3'') as SDSS fibers, suggesting no AGN is truly lineless. However, in no case does the line become more detectable as the aperture decreases, as would be expected if dilution by strong continuum was responsible for making the lines appear weak. We also show that the misclassified X-ray AGN retains the same position on the BPT diagram in smaller apertures (down to 0.''5), demonstrating that its misclassification is not the result of the dilution by HII regions. Thus, we conclude that continuum swamping or star formation dilution, i.e., aperture effects, are not responsible for atypical lines. Rather, the AGN lines are intrinsically weak.

astro-ph.GA↗

arXiv2022

A New Physical Picture for AGNs Lacking Optical Emission Lines

In this work, we use ~500 low-redshift (z ~ 0.1) X-ray AGNs observed by XMM-Newton and SDSS to investigate the prevalence and nature of AGNs that apparently lack optical emission lines (``optically dull AGNs''). Although 1/4 of spectra appear absorption-line dominated in visual assessment, line extraction with robust continuum subtraction from the MPA/JHU catalog reveals usable [OIII] measurements in 98% of the sample, allowing us to study [OIII]-underluminous AGNs together with more typical AGNs in the context of the L$_{\mathrm{[OIII]}}$--L$_{X}$ relation. We find that ``optically dull AGNs'' do not constitute a distinct population of AGNs. Instead, they are the [OIII]-underluminous tail of a single, unimodal L$_{\mathrm{[OIII]}}$--L$_{X}$ relation that has substantial scatter (0.6 dex). We find the degree to which an AGN is underluminous in [OIII] correlates with the specific SFR or D$_{4000}$ index of the host, which are both linked to the molecular gas fraction. Thus the emerging physical picture for the large scatter seems to involve the gas content of the narrow-line region. We find no significant role for previously proposed scenarios for the presence of optically dull AGNs, such as host dilution or dust obscuration. Despite occasionally weak lines in SDSS spectra, >80% of X-ray AGNs are identified as such with the BPT diagram. >90% are classified as AGNs based only on [NII]/H$α$, providing more complete AGN samples when [OIII] or H$β$ are weak. X-ray AGNs with LINER spectra obey essentially the same \lxo\ relation as Seyfert 2s, suggesting their line emission is produced by AGN activity.

astro-ph.GA↗

arXiv2021

Physical Drivers of Emission Line Diversity of SDSS Seyfert 2s and LINERs After Removal of Contributions by Star Formation

Ionization sources other than HII regions give rise to the right-hand branch in the standard ([NII]) BPT diagram, populated by Seyfert 2s and LINERs. However, because the majority of Seyfert/LINER hosts are star forming (SF), HII regions contaminate the observed lines to some extent, making it unclear if the position along the branch is merely due to various degrees of mixing between pure Seyfert/LINER and SF, or whether it reflects the intrinsic diversity of Seyfert/LINER ionizing sources. In this study, we empirically remove SF contributions in ~100,000 Seyfert/LINERs from SDSS using the doppelganger method. We find that mixing is not the principal cause of the extended morphology of the observed branch. Rather, Seyferts/LINERs intrinsically have a wide range of line ratios. Variations in ionization parameter and metallicity can account for much of the diversity of Seyfert/LINER line ratios, but the hardness of ionization field also varies significantly. Furthermore, our k-means classification on seven decontaminated emission lines reveals that LINERs are made up of two populations, which we call soft and hard LINERs. The Seyfert 2s differ from both types of LINERs primarily by higher ionization parameter, whereas the two LINER types mainly differ from each other (and from star-forming regions) in the hardness of the radiation field. We confirm that the [NII] BPT diagram more efficiently identifies LINERs than [SII] and [OI] diagnostics, because in the latter many LINERs, especially soft ones, occupy the same location as pure star-formers, even after the SF has been removed from LINER emission.

astro-ph.GA↗

arXiv2020

In-plane Néel wall chirality and orientation of interfacial Dzyaloshinskii-Moriya vector in magnetic films

The interfacial Dzyaloshinskii-Moriya interaction (DMI) is of great interest as it can stabilize chiral spin structures in thin films. Experiments verifying the orientation of the interfacial DMI vector remain rare, in part due to the difficulty of separating vector components of DMI. In this study, Fe/Ni bilayers and Co/Ni multilayers were deposited epitaxially onto Cu(001) and Pt(111) substrates, respectively. By tailoring the effective anisotropy, spin reorientation transitions (SRTs) are employed to probe the orientation of the DMI vector by measuring the spin structure of domain walls on both sides of the SRTs. The interfacial DMI is found to be sufficiently strong to stabilize chiral Néel walls in the out-of-plane magnetized regimes, while achiral Néel walls are observed in the in-plane magnetized regimes. These findings experimentally confirm that the out-of-plane component of the DMI vector is insignificant in these fcc(001) and fcc(111) oriented interfaces, even in the presence of atomic steps.

cond-mat.mtrl-sci↗

arXiv2019

Crossing the Line: Active Galactic Nuclei in the Star-forming region of the BPT Diagram

In this work, we investigate the reliability of the BPT diagram for excluding galaxies that host an AGN. We determine the prevalence of X-ray AGN in the star-forming region of the BPT diagram and discuss the reasons behind this apparent misclassification, focusing primarily on relatively massive ($\log(M_{*})\gtrsim10$) galaxies. X-ray AGN are selected from deep XMM observations using a new method that results in greater samples with a wider range of X-ray luminosities, complete to $\log(L_{X})>41$ for $z<0.3$. Taking X-ray detectability into account, we find the average fraction of X-ray AGN in the BPT star-forming branch is 2$\%$, suggesting the BPT diagram can provide a reasonably clean sample of star-forming galaxies. However, the X-ray selection is itself rather incomplete. At the tip of the AGN branch of the BPT diagram, the X-ray AGN fraction is only 14$\%$, which may have implications for studies that exclude AGN based only on X-ray observations. Interestingly, the X-ray AGN fractions are similar for Seyfert and LINER populations, consistent with LINERs being true AGN. We find that neither the star-formation dilution nor the hidden broad-line components can satisfactorily explain the apparent misclassification of X-ray AGN. On the other hand, $\sim40\%$ of all X-ray AGN have weak emission lines such that they cannot be placed on the BPT diagram at all and often have low specific SFRs. Therefore, the most likely explanation for "misclassified" X-ray AGN is that they have intrinsically weak AGN lines, and are only placeable on the BPT diagram when they tend to have high specific SFRs.

astro-ph.GA↗