Exploring Image-Text Alignment for Radio Galaxy Morphologies

Erica Lastufka; Mariia Drozdova; Svyatoslav Volosynovskiy

arxiv: 2607.06305 · v1 · pith:FDDT4C5A · submitted 2026-07-07 · astro-ph.IM

Exploring Image-Text Alignment for Radio Galaxy Morphologies

Erica Lastufka , Mariia Drozdova , Svyatoslav Volosynovskiy This is my paper

Reviewed by Pith T0 review T1 audit T2 compute T3 formal T4 kernel 2026-07-08 10:12 UTCglm-5.2pith:FDDT4C5A record.json open to challenge →

classification astro-ph.IM PACS 95.75.Mn95.80.+p

keywords alignmentimagescaptionsfine-tuninggalaxyradiocaption-basedcapture

0 comments

The pith

Text captions match images on radio galaxy classification

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper asks whether natural language descriptions of radio galaxy images can carry the same morphological information as the images themselves. Using the MiraBest dataset of Fanaroff-Riley Type I and Type II radio galaxies, the authors generate text captions with a commercial vision-language model (Gemini 2.5-Flash) using both a generic prompt and a domain-specific prompt designed to elicit expert morphological descriptions. They then feed images and captions through the SigLIP-2 vision-language model, with and without lightweight LoRA fine-tuning, and evaluate whether text embeddings alone can distinguish FR-I from FR-II galaxies as well as image embeddings can. The central finding is that linear-probe F1 scores for text embeddings (~0.90) match those for image embeddings (~0.90) on this binary classification task, suggesting that captions encode enough morphological signal to serve as a complement to image-based representations. However, instance-level retrieval metrics (Recall@1 of 0.01–0.07) remain near zero, meaning the model can separate broad morphological classes but cannot reliably match a specific caption to its specific image. Fine-tuning with LoRA sharpens class boundaries and improves local clustering but does not improve global image-text alignment; the sigmoid loss decreases without pulling paired image-text embeddings closer together. A secondary finding is that scientifically detailed captions only marginally outperform generic, sometimes incorrect ones — Gemini's default vocabulary already contains enough discriminative signal for the binary task.

Core claim

Text embeddings derived from model-generated captions of radio galaxy images achieve the same linear-probe F1 (~0.90) as image embeddings on FR-I/FR-II binary classification, but instance-level alignment between specific images and their captions remains essentially absent (Recall@1 of 0.01–0.07). Fine-tuning improves class separability but not cross-modal retrieval, revealing a gap between class-level discrimination and instance-level morphological fidelity.

What carries the argument

The central object is the SigLIP-2 vision-language model, which produces joint embedding spaces for images and text. The authors use linear probes (lightweight classifiers on frozen embeddings) to test whether each modality preserves enough information to separate FR-I from FR-II galaxies, and retrieval metrics (Recall@1, class-level Recall@1) to test whether specific image-text pairs can be matched. LoRA fine-tuning adapts a small fraction (0.16–0.33%) of model parameters to the radio galaxy domain.

If this is right

If text embeddings genuinely carry morphological signal, natural language captions could serve as a lightweight labeling layer for large radio surveys, complementing tabular source catalogs that capture only patch-level Gaussian fits.
The gap between class-level F1 and instance-level Recall@1 suggests that current caption-based embeddings encode coarse categorical information but not fine-grained morphological detail, limiting their use for similarity-based discovery or individual source characterization.
The finding that generic captions perform nearly as well as expert ones implies that the discriminative signal may come from broad vocabulary differences between FR-I and FR-II descriptions rather than from genuine morphological reasoning, which would narrow the scope of the claim.
If the approach scales to larger and more morphologically diverse datasets, caption-based embeddings could enable zero-shot or few-shot classification pipelines for upcoming surveys without requiring labeled training images.

Load-bearing premise

The paper treats binary FR-I/FR-II classification as a sufficient proxy for 'morphological information.' If the text embeddings merely encode which of two classes a galaxy belongs to — something a captioning model can infer from gross visual cues — rather than genuine morphological detail (lobe structure, jet orientation, diffuse emission patterns), then the claim that captions 'encode meaningful morphological information' is stronger than the evidence supports. The near-zero

What would settle it

If text embeddings continue to match image embeddings on binary classification but fail on any finer morphological task (e.g., multi-class morphology, regression of physical parameters, or instance retrieval), then the comparable F1 scores reflect class-label encoding rather than morphological information transfer.

Figures

Figures reproduced from arXiv: 2607.06305 by Erica Lastufka, Mariia Drozdova, Svyatoslav Volosynovskiy.

**Figure 2.** Figure 2: Examples of FR-I, FR-II, FR-I, and FR-II galaxies from the MiraBest [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗

**Figure 3.** Figure 3: Cosine similarity between image and curated text embeddings for the [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗

**Figure 4.** Figure 4: T-SNE visualization of text or features extracted from the frozen [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗

**Figure 5.** Figure 5: T-SNE visualization of text or image features extracted after various [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗

read the original abstract

We investigate whether specially constructed text captions can capture the same morphological information as radio galaxy images. Using the MiraBest dataset, we generate captions with a domain-specific prompt and evaluate their alignment with images through the SigLIP-2 vision--language model, with and without LoRA fine-tuning. Results show that caption-based classification of FR-I and FR-II galaxies performs similarly to images, with fine-tuning improving local coherence of embeddings but not global alignment.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Text captions match images on binary FR-I/FR-II F1, but the evidence points to class-correlated vocabulary, not morphological encoding.

read the letter

Short version: this paper asks a genuinely new question — can text captions encode radio galaxy morphology the way images do? — and the answer it actually provides is narrower than the one it claims. The application of SigLIP-2 to radio galaxy captions on MiraBest is a legitimate first look at a problem nobody else has tackled in this domain. The authors also deserve credit for reporting multiple metrics that, read together, undermine their own headline claim. That honesty matters. The systematic comparison of curated vs. control captions is a good experimental design choice, and the t-SNE plots and cosine similarity analysis in the appendices are useful supporting material. The finding that LoRA fine-tuning sharpens class boundaries without improving global image-text alignment is a real and interesting observation about how contrastive loss behaves on small scientific datasets. Here is the problem. The central claim — that captions “encode meaningful morphological information” — does not survive the paper's own Table 1. Control captions, which the authors describe as having incorrect descriptors and less morphological detail, achieve essentially the same linear probe F1 as curated captions (differences of 0.01–0.03). The authors call this “surprising,” but it is actually the key result: the F1 is measuring whether Gemini's vocabulary happens to correlate with the binary label, not whether the text encodes morphology. Two other metrics in the same table confirm this. Recall@1 is near zero (0.01–0.07), meaning text embeddings cannot retrieve their paired image. KNN F1 for text (0.69–0.73) is well below images (0.81–0.88), meaning text embeddings do not form locally coherent morphological neighborhoods. So what the paper actually shows is that text embeddings carry enough class-level signal for a linear probe to separate two categories on 104 test samples — not that they encode morphological information in any meaningful sense. The manual editing of test-set captions is a second concern: a human editor aware of the FR-I/FR-II distinction may inject class-discriminative vocabulary, which would inflate text F1 independently of any morphological content. No error bars are reported on any metric, and with 104 test samples, the 0.01–0.03 differences between curated and control captions are almost certainly noise. No code or data is shipped, and the captioning model (Gemini) is proprietary, so the pipeline is not independently reproducible. This is a reasonable proof-of-concept by people asking the right question. The conclusions need to be scaled back to match what the metrics actually show: binary class separability, not morphological encoding. A serious referee should push the authors to (1) report confidence intervals, (2) use unedited test captions, (3) test on a larger dataset, and (4) reframe the claim. The paper is worth refereeing because the application is novel and the empirical work, while limited, is transparent enough to critique productively.

Referee Report

3 major / 8 minor

Summary. This manuscript investigates whether text captions generated by a large language model (Gemini 2.5-Flash) can encode morphological information about radio galaxies comparable to image-based representations. Using the MiraBest dataset (833 training, 104 test images), the authors generate captions with both a domain-specific prompt and a generic control prompt, then evaluate image-text alignment through the SigLIP-2 vision-language model with and without LoRA fine-tuning. The central finding is that linear-probe F1 scores for text embeddings (~0.90) match those for image embeddings (~0.90) on binary FR-I/FR-II classification, leading the authors to conclude that captions 'encode meaningful morphological information' and are a 'viable complement' to image representations. The paper also reports that fine-tuning improves local coherence but not global alignment, and that control captions perform nearly as well as curated ones.

Significance. The question of whether natural language captions can serve as a useful modality for astronomical data is timely and relevant to the astro-ph.IM community. The systematic comparison of curated vs. control captions, multiple fine-tuning configurations, and a suite of metrics (linear probe, KNN, retrieval) is a reasonable experimental design. The honest reporting of near-zero Recall@1 and the surprising parity between curated and control captions are commendable, as these findings are informative even though they complicate the paper's narrative. However, the significance is limited by the small test set (104 samples), the binary classification task that reduces morphology to two classes, and the absence of error bars or confidence intervals on any reported metric. The claim that captions encode 'meaningful morphological information' is not adequately supported by the evidence presented, as detailed below.

major comments (3)

§3, Table 1: No error bars or confidence intervals are reported on any metric. With a test set of only 104 samples, the differences between curated and control captions (shown in parentheses, typically ±0.01–0.03 F1) are well within expected statistical fluctuation. Without bootstrap confidence intervals or similar uncertainty estimates, it is impossible to determine whether any reported difference is meaningful. This is load-bearing because the paper's narrative relies on interpreting these small differences (e.g., 'scientific captions only provide a small boost in performance'). Adding bootstrap CIs would either confirm or refute this interpretation.
§4, Discussion: The paper's central claim that captions 'encode meaningful morphological information' is contradicted by its own evidence. Table 1 shows that control captions—which the authors describe as having 'less morphological detail and often assumed incorrect descriptors' (§2)—achieve essentially the same linear-probe F1 as curated captions. The authors acknowledge this is 'surprising' but do not reconcile it with their conclusion. If incorrect captions perform as well as detailed ones, the F1 is measuring class-correlated vocabulary, not morphological information content. This is further supported by near-zero Recall@1 (0.01–0.07), indicating text embeddings lack instance-level morphological detail, and by KNN F1 for text (0.69–0.73) being substantially lower than for images (0.81–0.88). The conclusion should be revised to accurately reflect what the metrics show: text embeddings
§2, Data: The curated captions for the test set were 'manually controlled, and if need be, edited' by the authors. This introduces potential evaluator bias, as the same individuals who designed the experiment curated the test-set captions. While the paper notes this was done only for the test set, no details are given on the extent or criteria for editing. A brief description of the editing protocol, or an acknowledgment of this as a limitation, would strengthen the manuscript's credibility.

minor comments (8)

Table 1: The parenthetical differences (e.g., '(-0.01)', '(+0.03)') are ambiguous—do they represent curated minus control, or control minus curated? A footnote or caption clarification would improve readability.
§3: The validation subset of ~100 samples from the training set is small. The sensitivity of early stopping to this choice should be briefly discussed.
Appendix C: The metric definitions are helpful but could note that Recall@1 is computed over the test set (not the full dataset) to avoid ambiguity.
§4: The phrase 'asymmetrical semantic anchoring' is introduced without sufficient definition. A clearer explanation of this concept would help readers.
Figure 1: Axis labels and legends could be more explicit about which embeddings are being compared.
§1: The reference to 'AstroLLaMA' appears to have a formatting issue (missing space or citation delimiter).
§2: The Gemini model version string is very specific; a note on reproducibility given that API versions change would be appropriate.
The paper would benefit from a brief discussion of potential failure modes or biases when applying this approach to larger, less curated datasets (e.g., with artifacts, noise, or hybrid sources).

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for a careful and constructive reading of our manuscript. The report correctly identifies several genuine weaknesses—particularly the absence of uncertainty estimates and an overstatement of our central claim—that we will address in revision. We respond to each major comment below.

read point-by-point responses

Referee: §3, Table 1: No error bars or confidence intervals are reported on any metric. With a test set of only 104 samples, the differences between curated and control captions are well within expected statistical fluctuation. Without bootstrap confidence intervals or similar uncertainty estimates, it is impossible to determine whether any reported difference is meaningful.

Authors: The referee is correct. With 104 test samples, the differences reported in Table 1 (typically ±0.01–0.03 F1) are well within the range of statistical fluctuation, and we should not have drawn interpretive conclusions from them without uncertainty estimates. We will add bootstrap confidence intervals (1000 resamples, 95% CI) to all metrics in Table 1. This will allow readers to assess which differences are meaningful and which are not. We expect this will confirm that the curated-vs-control differences are not statistically significant, which is consistent with the narrative we already report (that the parity is 'surprising'). However, we agree that the current presentation does not support this interpretation quantitatively, and the revision will fix this. revision: yes
Referee: §4, Discussion: The paper's central claim that captions 'encode meaningful morphological information' is contradicted by its own evidence. Control captions achieve essentially the same linear-probe F1 as curated captions. If incorrect captions perform as well as detailed ones, the F1 is measuring class-correlated vocabulary, not morphological information content. Near-zero Recall@1 and lower KNN F1 for text further undermine the claim. The conclusion should be revised to accurately reflect what the metrics show.

Authors: We largely accept this criticism. The referee is right that the parity between curated and control captions, combined with near-zero Recall@1 and substantially lower KNN F1 for text (0.69–0.73 vs. 0.81–0.88 for images), complicates the claim that captions encode 'meaningful morphological information.' The linear-probe F1 of ~0.90 for text likely reflects class-correlated vocabulary rather than instance-level morphological detail, as the referee suggests. We will revise the conclusion to state more precisely what the evidence supports: text embeddings carry class-level discriminative signal (sufficient for binary FR-I/FR-II separation) but do not preserve instance-level morphological detail (as shown by near-zero Recall@1 and lower KNN F1). The phrase 'meaningful morphological information' will be replaced with more precise language, and the claim that captions are a 'viable complement' to image representations will be qualified to refer specifically to class-level labeling rather than instance-level morphological encoding. We retain the point that class-level Recall@1 (0.53–0.77) does indicate some morphological structure in the text embedding space, but we will frame this more cautiously. revision: yes
Referee: §2, Data: The curated captions for the test set were 'manually controlled, and if need be, edited' by the authors, introducing potential evaluator bias. No details are given on the extent or criteria for editing. A brief description of the editing protocol, or an acknowledgment of this as a limitation, would strengthen the manuscript's credibility.

Authors: This is a fair point. We will add a description of the editing protocol: the test-set captions were reviewed by one author (EL), with corrections limited to fixing factually incorrect descriptors (e.g., wrong spatial locations of features, incorrect morphological terms) and removing speculative language about the nature of emission. No captions were rewritten wholesale; the goal was to ensure factual accuracy of morphological description, not to optimize for classification performance. We will also add an explicit acknowledgment of evaluator bias as a limitation, noting that an independent annotator or blinded protocol would be preferable for future work at scale. We agree this transparency is important for credibility. revision: yes

Circularity Check

0 steps flagged

No significant circularity; the paper is an empirical ML evaluation with no derivation chain that reduces to its inputs by construction.

full rationale

The paper's central claim — that caption-based embeddings encode morphological information — is supported by standard ML metrics (linear probe F1, KNN F1, Recall@1) computed on a held-out test set (Table 1). There is no equation or derivation chain where a predicted quantity is defined in terms of the quantity it claims to predict. The F1 scores are not fitted parameters renamed as predictions; they are evaluation metrics on test data. The one self-citation (Lastufka et al. 2024, with author overlap) is cited as prior-work context for self-supervised learning on MeerKAT images and is not load-bearing for the central claim, which rests on Table 1 results. The concern that test-set captions were 'manually controlled, and if need be, edited' (§2) is a data-contamination risk, not a circularity in the derivation. The observation that control captions achieve similar F1 to curated ones — which the authors themselves flag as 'surprising' — is a validity concern about whether binary F1 measures morphological information, but it is not circular: the metrics are computed independently of any fitted parameter that would force the result. No uniqueness theorem, no ansatz smuggled via self-citation, no renaming of a known result. The paper is self-contained against external benchmarks (MiraBest is a public dataset, SigLIP-2 is a public model). Score 1 reflects the minor, non-load-bearing self-citation.

Axiom & Free-Parameter Ledger

5 free parameters · 4 axioms · 0 invented entities

The paper introduces no new physical entities, particles, forces, or mathematical objects. The free parameters are standard ML hyperparameters. The axioms are domain assumptions about the suitability of the task proxy, the captioning model, the embedding model, and the manual editing procedure.

free parameters (5)

LoRA rank R = 8
Chosen hyperparameter for low-rank adaptation; not tuned via systematic search as far as the paper reports.
LoRA alpha = 16
Chosen hyperparameter; standard value but not justified by systematic search.
Learning rate = 2.5e-5
Chosen hyperparameter for fine-tuning; no search or justification provided.
KNN k value = 5
Chosen for KNN classifier; footnote states 'We chose k=5' without justification.
Validation subset size = ~100
Used for early stopping; size chosen without stated justification.

axioms (4)

domain assumption Binary FR-I/FR-II classification is a sufficient proxy for 'morphological information' in radio galaxies.
Invoked implicitly throughout §3 and §4; the paper's headline claim about encoding morphological information rests on binary classification F1 as the primary metric.
domain assumption Gemini 2.5-Flash can generate morphologically accurate captions from radio galaxy images.
Invoked in §2 where captions are generated; the quality of the entire pipeline depends on this assumption, which is only partially validated by manual inspection of training captions.
domain assumption SigLIP-2's pre-trained joint embedding space is suitable for radio galaxy images despite being trained on natural images.
Invoked in §3 where SigLIP-2 is used without pre-training on astronomical data; no domain adaptation of the base model is performed before LoRA.
ad hoc to paper Manual editing of test-set captions does not introduce systematic bias.
Invoked in §2: 'curated captions were manually controlled, and if need be, edited for the test set only.' This is a design choice that affects evaluation integrity.

pith-pipeline@v1.1.0-glm · 10071 in / 2935 out tokens · 472109 ms · 2026-07-08T10:12:44.165719+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

19 extracted references · 19 canonical work pages · 6 internal anchors

[1]

Self-supervised

Hayat, Md Abul and Stein, George and Harrington, Peter and Lukić, Zarija and Mustafa, Mustafa , month = apr, year =. Self-supervised. ApJL , publisher =. doi:10.3847/2041-8213/abf2c7 , abstract =

work page doi:10.3847/2041-8213/abf2c7 2041
[2]

RAS Techniques and Instruments , author =

Radio galaxy zoo: towards building the first multipurpose foundation model for radio astronomy with self-supervised learning , volume =. RAS Techniques and Instruments , author =. 2024 , pages =. doi:10.1093/rasti/rzad055 , abstract =

work page doi:10.1093/rasti/rzad055 2024
[3]

2023 , pages =

RAS Techniques and Instruments , author =. 2023 , pages =. doi:10.1093/rasti/rzad017 , abstract =

work page doi:10.1093/rasti/rzad017 2023
[4]

Monthly Notices of the Royal Astronomical Society , author =

Enabling unsupervised discovery in astronomical images through self-supervised representations , volume =. Monthly Notices of the Royal Astronomical Society , author =. 2024 , pages =. doi:10.1093/mnras/stae926 , abstract =

work page doi:10.1093/mnras/stae926 2024
[5]

2024 , keywords =

Publications of the Astronomical Society of Australia , author =. 2024 , keywords =. doi:10.1017/pasa.2023.64 , abstract =

work page doi:10.1017/pasa.2023.64 2024
[6]

The information bottleneck method

Tishby, Naftali and Pereira, Fernando C. and Bialek, William , month = apr, year =. The information bottleneck method , url =. Proc. of the 37-th. doi:10.48550/arXiv.physics/0004057 , abstract =

work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.physics/0004057
[7]

DINOv2 Meets Text: A Unified Framework for Image- and Pixel-Level Vision-Language Alignment

Jose, Cijo and Moutakanni, Théo and Kang, Dahyun and Baldassarre, Federico and Darcet, Timothée and Xu, Hu and Li, Daniel and Szafraniec, Marc and Ramamonjisoa, Michaël and Oquab, Maxime and Siméoni, Oriane and Vo, Huy V. and Labatut, Patrick and Bojanowski, Piotr , month = dec, year =. doi:10.48550/arXiv.2412.16334 , abstract =

work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2412.16334
[8]

radio-llava: Advancing Vision-Language Models for Radio Astronomical Source Analysis

Riggi, S. and Cecconello, T. and Pilzer, A. and Palazzo, S. and Gupta, N. and Hopkins, A. M. and Trigilio, C. and Umana, G. , month = apr, year =. Evaluating small vision-language models as. doi:10.48550/arXiv.2503.23859 , abstract =

work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2503.23859
[9]

Monthly Notices of the Royal Astronomical Society , author =

Data-efficient classification of radio galaxies , volume =. Monthly Notices of the Royal Astronomical Society , author =. 2022 , pages =. doi:10.1093/mnras/stab3144 , abstract =

work page doi:10.1093/mnras/stab3144 2022
[10]

CosmoCLIP: Generalizing Large Vision-Language Models for Astronomical Imaging

Imam, Raza and Alam, Mohammed Talha and Rahman, Umaima and Guizani, Mohsen and Karray, Fakhri , month = nov, year =. doi:10.48550/arXiv.2407.07315 , abstract =

work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2407.07315
[11]

doi:10.1109/SCW63240.2024.00019 , abstract =

Pan, Rui and Nguyen, Tuan Dung and Arora, Hardik and Accomazzi, Alberto and Ghosal, Tirthankar and Ting, Yuan-Sen , month = nov, year =. doi:10.1109/SCW63240.2024.00019 , abstract =

work page doi:10.1109/scw63240.2024.00019 2024
[12]

Zaman, Sharaf and Smith, Michael J. and Khetarpal, Pranav and Chakrabarty, Rishabh and Ginolfi, Michele and Huertas-Company, Marc and Jabłońska, Maja and Kruk, Sandor and Lain, Matthieu Le and Méndez, Sergio José Rodríguez and Tanoglidis, Dimitrios , month = apr, year =. doi:10.48550/arXiv.2504.08583 , abstract =

work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2504.08583
[13]

and Shen, Yelong and Wallis, Phillip and Allen-Zhu, Zeyuan and Li, Yuanzhi and Wang, Shean and Wang, Lu and Chen, Weizhu , month = oct, year =

Hu, Edward J. and Shen, Yelong and Wallis, Phillip and Allen-Zhu, Zeyuan and Li, Yuanzhi and Wang, Shean and Wang, Lu and Chen, Weizhu , month = oct, year =

work page
[14]

and Bloom, J

Rizhko, M. and Bloom, J. S. , month = jun, year =. AJ , publisher =. doi:10.3847/1538-3881/adcbad , abstract =

work page doi:10.3847/1538-3881/adcbad
[15]

SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Features

Tschannen, Michael and Gritsenko, Alexey and Wang, Xiao and Naeem, Muhammad Ferjad and Alabdulmohsin, Ibrahim and Parthasarathy, Nikhil and Evans, Talfan and Beyer, Lucas and Xia, Ye and Mustafa, Basil and Hénaff, Olivier and Harmsen, Jeremiah and Steiner, Andreas and Zhai, Xiaohua , month = feb, year =. doi:10.48550/arXiv.2502.14786 , abstract =

work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2502.14786
[16]

and Bui, Thang and Goodman, Alyssa and Accomazzi, Alberto and Naiman, Jill and Cranney, Jesse and Schawinski, Kevin and Raileanu, Roberta , editor =

Nguyen, Tuan Dung and Ting, Yuan-Sen and Ciuca, Ioana and O'Neill, Charles and Sun, Ze-Chang and Jabłońska, Maja and Kruk, Sandor and Perkowski, Ernest and Miller, Jack and Li, Jason Jason Jingsh and Peek, Josh and Iyer, Kartheik and Rozanski, Tomasz and Khetarpal, Pranav and Zaman, Sharaf and Brodrick, David and Rodriguez Mendez, Sergio J. and Bui, Thang...

work page doi:10.18653/v1/2023.wiesp-1.7 2023
[17]

and Sourav, Shashwat and Wu, John F

UniverseTBD and Duraphe, Kshitij and Smith, Michael J. and Sourav, Shashwat and Wu, John F. , month = sep, year =. The. doi:10.48550/arXiv.2509.19453 , abstract =

work page doi:10.48550/arxiv.2509.19453
[18]

and Bait, O

Lastufka, E. and Bait, O. and Drozdova, M. and Kinakh, V. and Piras, D. and Audard, M. and Dessauges-Zavadsky, M. and Holotyak, T. and Schaerer, D. and Voloshynovskiy, S. , month = nov, year =. Examining vision foundation models for classification and detection in optical and radio astronomy , volume =. A&A , publisher =. doi:10.1051/0004-6361/202553691 ,...

work page doi:10.1051/0004-6361/202553691
[19]

and Bait, O

Lastufka, E. and Bait, O. and Taran, O. and Drozdova, M. and Kinakh, V. and Piras, D. and Audard, M. and Dessauges-Zavadsky, M. and Holotyak, T. and Schaerer, D. and Voloshynovskiy, S. , month = oct, year =. Self-supervised learning on. A&A , publisher =. doi:10.1051/0004-6361/202449964 , abstract =

work page doi:10.1051/0004-6361/202449964

[1] [1]

Self-supervised

Hayat, Md Abul and Stein, George and Harrington, Peter and Lukić, Zarija and Mustafa, Mustafa , month = apr, year =. Self-supervised. ApJL , publisher =. doi:10.3847/2041-8213/abf2c7 , abstract =

work page doi:10.3847/2041-8213/abf2c7 2041

[2] [2]

RAS Techniques and Instruments , author =

Radio galaxy zoo: towards building the first multipurpose foundation model for radio astronomy with self-supervised learning , volume =. RAS Techniques and Instruments , author =. 2024 , pages =. doi:10.1093/rasti/rzad055 , abstract =

work page doi:10.1093/rasti/rzad055 2024

[3] [3]

2023 , pages =

RAS Techniques and Instruments , author =. 2023 , pages =. doi:10.1093/rasti/rzad017 , abstract =

work page doi:10.1093/rasti/rzad017 2023

[4] [4]

Monthly Notices of the Royal Astronomical Society , author =

Enabling unsupervised discovery in astronomical images through self-supervised representations , volume =. Monthly Notices of the Royal Astronomical Society , author =. 2024 , pages =. doi:10.1093/mnras/stae926 , abstract =

work page doi:10.1093/mnras/stae926 2024

[5] [5]

2024 , keywords =

Publications of the Astronomical Society of Australia , author =. 2024 , keywords =. doi:10.1017/pasa.2023.64 , abstract =

work page doi:10.1017/pasa.2023.64 2024

[6] [6]

The information bottleneck method

Tishby, Naftali and Pereira, Fernando C. and Bialek, William , month = apr, year =. The information bottleneck method , url =. Proc. of the 37-th. doi:10.48550/arXiv.physics/0004057 , abstract =

work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.physics/0004057

[7] [7]

DINOv2 Meets Text: A Unified Framework for Image- and Pixel-Level Vision-Language Alignment

Jose, Cijo and Moutakanni, Théo and Kang, Dahyun and Baldassarre, Federico and Darcet, Timothée and Xu, Hu and Li, Daniel and Szafraniec, Marc and Ramamonjisoa, Michaël and Oquab, Maxime and Siméoni, Oriane and Vo, Huy V. and Labatut, Patrick and Bojanowski, Piotr , month = dec, year =. doi:10.48550/arXiv.2412.16334 , abstract =

work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2412.16334

[8] [8]

radio-llava: Advancing Vision-Language Models for Radio Astronomical Source Analysis

Riggi, S. and Cecconello, T. and Pilzer, A. and Palazzo, S. and Gupta, N. and Hopkins, A. M. and Trigilio, C. and Umana, G. , month = apr, year =. Evaluating small vision-language models as. doi:10.48550/arXiv.2503.23859 , abstract =

work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2503.23859

[9] [9]

Monthly Notices of the Royal Astronomical Society , author =

Data-efficient classification of radio galaxies , volume =. Monthly Notices of the Royal Astronomical Society , author =. 2022 , pages =. doi:10.1093/mnras/stab3144 , abstract =

work page doi:10.1093/mnras/stab3144 2022

[10] [10]

CosmoCLIP: Generalizing Large Vision-Language Models for Astronomical Imaging

Imam, Raza and Alam, Mohammed Talha and Rahman, Umaima and Guizani, Mohsen and Karray, Fakhri , month = nov, year =. doi:10.48550/arXiv.2407.07315 , abstract =

work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2407.07315

[11] [11]

doi:10.1109/SCW63240.2024.00019 , abstract =

Pan, Rui and Nguyen, Tuan Dung and Arora, Hardik and Accomazzi, Alberto and Ghosal, Tirthankar and Ting, Yuan-Sen , month = nov, year =. doi:10.1109/SCW63240.2024.00019 , abstract =

work page doi:10.1109/scw63240.2024.00019 2024

[12] [12]

Zaman, Sharaf and Smith, Michael J. and Khetarpal, Pranav and Chakrabarty, Rishabh and Ginolfi, Michele and Huertas-Company, Marc and Jabłońska, Maja and Kruk, Sandor and Lain, Matthieu Le and Méndez, Sergio José Rodríguez and Tanoglidis, Dimitrios , month = apr, year =. doi:10.48550/arXiv.2504.08583 , abstract =

work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2504.08583

[13] [13]

and Shen, Yelong and Wallis, Phillip and Allen-Zhu, Zeyuan and Li, Yuanzhi and Wang, Shean and Wang, Lu and Chen, Weizhu , month = oct, year =

Hu, Edward J. and Shen, Yelong and Wallis, Phillip and Allen-Zhu, Zeyuan and Li, Yuanzhi and Wang, Shean and Wang, Lu and Chen, Weizhu , month = oct, year =

work page

[14] [14]

and Bloom, J

Rizhko, M. and Bloom, J. S. , month = jun, year =. AJ , publisher =. doi:10.3847/1538-3881/adcbad , abstract =

work page doi:10.3847/1538-3881/adcbad

[15] [15]

SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Features

Tschannen, Michael and Gritsenko, Alexey and Wang, Xiao and Naeem, Muhammad Ferjad and Alabdulmohsin, Ibrahim and Parthasarathy, Nikhil and Evans, Talfan and Beyer, Lucas and Xia, Ye and Mustafa, Basil and Hénaff, Olivier and Harmsen, Jeremiah and Steiner, Andreas and Zhai, Xiaohua , month = feb, year =. doi:10.48550/arXiv.2502.14786 , abstract =

work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2502.14786

[16] [16]

and Bui, Thang and Goodman, Alyssa and Accomazzi, Alberto and Naiman, Jill and Cranney, Jesse and Schawinski, Kevin and Raileanu, Roberta , editor =

Nguyen, Tuan Dung and Ting, Yuan-Sen and Ciuca, Ioana and O'Neill, Charles and Sun, Ze-Chang and Jabłońska, Maja and Kruk, Sandor and Perkowski, Ernest and Miller, Jack and Li, Jason Jason Jingsh and Peek, Josh and Iyer, Kartheik and Rozanski, Tomasz and Khetarpal, Pranav and Zaman, Sharaf and Brodrick, David and Rodriguez Mendez, Sergio J. and Bui, Thang...

work page doi:10.18653/v1/2023.wiesp-1.7 2023

[17] [17]

and Sourav, Shashwat and Wu, John F

UniverseTBD and Duraphe, Kshitij and Smith, Michael J. and Sourav, Shashwat and Wu, John F. , month = sep, year =. The. doi:10.48550/arXiv.2509.19453 , abstract =

work page doi:10.48550/arxiv.2509.19453

[18] [18]

and Bait, O

Lastufka, E. and Bait, O. and Drozdova, M. and Kinakh, V. and Piras, D. and Audard, M. and Dessauges-Zavadsky, M. and Holotyak, T. and Schaerer, D. and Voloshynovskiy, S. , month = nov, year =. Examining vision foundation models for classification and detection in optical and radio astronomy , volume =. A&A , publisher =. doi:10.1051/0004-6361/202553691 ,...

work page doi:10.1051/0004-6361/202553691

[19] [19]

and Bait, O

Lastufka, E. and Bait, O. and Taran, O. and Drozdova, M. and Kinakh, V. and Piras, D. and Audard, M. and Dessauges-Zavadsky, M. and Holotyak, T. and Schaerer, D. and Voloshynovskiy, S. , month = oct, year =. Self-supervised learning on. A&A , publisher =. doi:10.1051/0004-6361/202449964 , abstract =

work page doi:10.1051/0004-6361/202449964