Datasets:

HGTasd
/

PosterText-30K

instruction stringclasses 95 values	input stringlengths 141 265k	output stringlengths 15 1.61k
Generate 2 bullet points for the "Experimental Results / Performance Analysis" section of an academic poster.	## Paper Content # Paper Title Simpler Diffusion: 1.5 FID on ImageNet512 with pixel-space diffusion ## Abstract Latent diffusion models have become the popular choice for scaling up diffusion models for high resolution image synthesis. Compared to pixel-space models that are trained end-to-end, latent models are pe...	• Compares FID scores on MSCOCO dataset • SiD2 achieves 8.1 FID with bias=-3 and 6.7 FID with 16-step distilled version
Generate 4 bullet points for the "Core Method / Technical Approach" section of an academic poster.	## Paper Content # Paper Title TWO EFFECTS, ONE TRIGGER: ON THE MODALITY GAP, OBJECT BIAS, AND INFORMATION IMBALANCE IN CONTRASTIVE VISION-LANGUAGE MODELS ## Abstract Contrastive vision-language models (VLMs), like CLIP, have gained popularity for their versatile applicability to various downstream tasks. Despite t...	• How to measure modality gap: RMG formula based on pairwise distances • Only a few embedding dimensions drive the modality gap • Removal of those dimensions doesn't improve performance • Other factors (like model size) are more important to performance
Generate 3 bullet points for the "Core Method / Technical Approach" section of an academic poster.	## Paper Content # Paper Title Learning Trimodal Relation for Audio-Visual Question Answering with Missing Modality ## Method Fig. 2 shows the overall architecture of the proposed AVQA framework addressing missing modalities during inference. The visual modal input, audio input, and question pass through their corr...	• Audio-Visual Relation-aware Diffusion Model: Enhances feature representation of the missed modality. • Diffusion process defined as a Gaussian noise removal process over timesteps. • Uses weight sharing and AVQA backbone for final prediction.
Generate 3 bullet points for the "Core Method / Technical Approach" section of an academic poster.	## Paper Content # Paper Title SocraticLM: Exploring Socratic Personalized Teaching with Large Language Models ## Abstract Large language models (LLMs) are considered a crucial technology for advancing intelligent education since they exhibit the potential for an in-depth understanding of teaching scenarios and pro...	• Proposes five metrics to evaluate LLM teaching quality: Overall Quality (Overall), Incorrect Answer Recognition Accuracy (IAR4), Correct Answer Recognition Accuracy (CAR4), Successful Explanation Rate (SER), and Successful Rejection Rate (SRR). • Motivation: previous metrics calculate similarity between model-generat...
Generate 3 bullet points for the "Background / Related Work" section of an academic poster.	## Paper Content # Paper Title Towards Optimizing Large-Scale Multi-Graph Matching in Bioimaging ## Abstract Multi-graph matching is an important problem in computer vision. Our task comes from bioimaging, where a set of 100 3D-microscopic images of worms have to be brought into correspondence. Surprisingly, virtua...	• Theoretical lower bound: incomplete multi-graph matching is O(n·d) • Complete multi-graph matching is O(n·d²) • Solvers must handle incomplete problems directly to be efficient
Generate 1 bullet points for the "Other Content" section of an academic poster.	## Paper Content # Paper Title Attention Calibration for Disentangled Text-to-Image Personalization ## Abstract Recent thrilling progress in large-scale text-to-image (T2I) models has unlocked unprecedented synthesis quality of AI-generated content (AIGC) including image generation, 3D and video composition. Furthe...	• Datasets and code are publicly available at: https://github.com/Monalissaa/DisenDiff
Generate 3 bullet points for the "Core Method / Technical Approach" section of an academic poster.	## Paper Content # Paper Title Tiny Models are the Computational Saver for Large Models ## Method In EE-based methods, whether to exit after a specific layer or not can be controlled by many factors. Confidence-based exiting is one of the easiest to implement. Here, EEs are designed to fulfil two primary functions:...	• By reducing the exiting threshold t, we can find how many samples S can take without hurting overall performance. • r_match = max_t { # { m : max_i(y_S^(m,i)) ≥ t } / M } subject to correct classification. • Reduction of complexity ΔC = E[C_TS] - C_B = (C_S / C_B) - r_match.
Generate 4 bullet points for the "Research Motivation / Problem Background" section of an academic poster.	## Paper Content # Paper Title Bones Can't Be Triangles: Accurate and Efficient Vertebrae Keypoint Estimation through Collaborative Error Revision ## Introduction Accurate vertebrae keypoint estimation from X-ray images is crucial for effective medical diagnosis and treatment planning [20, 23], with errors having s...	• Objective: Accurate and efficient vertebrae keypoint estimation in X-ray images with minimal user intervention. • Challenge: Similarity between vertebrae often leads to inevitable errors. • Manual corrections are time-consuming and labor-intensive. • Interactive keypoint estimation relies on user feedback to refine e...
Generate 3 bullet points for the "Conclusion / Future Work" section of an academic poster.	## Paper Content # Paper Title FedAWA: Adaptive Optimization of Aggregation Weights in Federated Learning Using Client Vectors ## Abstract Federated Learning (FL) has emerged as a promising framework for distributed machine learning, enabling collaborative model training without sharing local data, thereby preservi...	• Client vectors in federated learning effectively capture relevant information about local datasets. • The aggregated global vector is more closely aligned with the ideal update direction in federated learning (the direction that would be obtained under centralized training). • FedAWA adaptively adjusts aggregation we...
Generate 3 bullet points for the "Qualitative Results / Visualization" section of an academic poster.	## Paper Content # Paper Title Bi-directional Contextual Attention for 3D Dense Captioning ## Experiments Datasets. We focus on 3D dense captioning, leveraging two benchmark datasets: ScanRefer [6] and $\mathrm{Nr3D}$ [1]. These datasets offer an extensive human-generated description of 3D scenes and objects. Sca...	• Shows qualitative examples of generated captions for different methods (Scan2Cap, 3DVG-Transformer, VideoCap3D, BiCA) on ScanRefer. • BiCA generates more accurate and detailed captions, especially for spatial relationships and object attributes. • Failures due to low IoU are highlighted in red boxes.
Generate 4 bullet points for the "Core Method / Technical Approach" section of an academic poster.	## Paper Content # Paper Title Query Efficient Black-Box Visual Prompting with Subspace Learning ## Abstract Visual Prompt Learning (VPL) has emerged as a powerful strategy for harnessing the capabilities of large-scale pretrained models (PTMs) to tackle specific downstream tasks. However, the opaque nature of PTMs...	• The objective is to minimize downstream task loss by adapting frozen PTM via input space prompt optimization. • Inspired by low-dimensional intrinsic subspace methods, we propose a subspace learning based black-box visual prompt approach. • Learn projection matrices Âp and Âw by iteratively generating columns and con...
Generate 3 bullet points for the "Qualitative Results / Visualization" section of an academic poster.	## Paper Content # Paper Title Communication-Efficient Collaborative Perception via Information Filling with Codebook ## Abstract Collaborative perception empowers each agent to improve its perceptual ability through the exchange of perceptual messages with other agents. It inherently results in a fundamental trade...	• Visualizes ego detections, ego features, information scores, and collaborator scores. • Shows selection matrices and filled confidence maps. • Demonstrates collaborative detections with bounding boxes overlaid on point clouds.
Generate 3 bullet points for the "Other Content" section of an academic poster.	## Paper Content # Paper Title Hybrid Proposal Refiner: Revisiting DETR Series from the Faster R-CNN Perspective ## Abstract With the transformative impact of the Transformer, DETR pioneered the application of the encoder-decoder architecture to object detection. A collection of follow-up research, e.g., Deformable...	• Contact email: jzha0100@uni.sydney.edu.au • Please consider citing our work if you find it helpful. • QR codes provided for code and paper access.
Generate 3 bullet points for the "Research Motivation / Problem Background" section of an academic poster.	## Paper Content # Paper Title Leveraging Frame Affinity for sRGB-to-RAW Video De-rendering ## Abstract Unprocessed RAW video has shown distinct advantages over sRGB video in video editing and computer vision tasks. However, capturing RAW video is challenging due to limitations in bandwidth and storage. Various met...	• Unprocessed RAW video has advantages over sRGB in editing and vision tasks. • Capturing RAW video is challenging due to bandwidth and storage limitations. • Previous sRGB-to-RAW methods store per-image metadata; this work uses frame affinity instead.
Generate 3 bullet points for the "Core Method / Technical Approach" section of an academic poster.	## Paper Content # Paper Title NVComposer: Boosting Generative Novel View Synthesis with Multiple Sparse and Unposed Images ## Abstract Recent advancements in generative models have significantly improved novel view synthesis (NVS) from multi-view data. However, existing methods depend on external multiview alignme...	• Proposes a dual-stream diffusion model that generates image-pose bundles jointly, based on video prior. • Introduces a geometry-aware feature alignment adapter that distills geometric prior during training. • Architecture includes CLIP encoder, latent encoder/decoder, and dual streams for image and pose generation.
Generate 4 bullet points for the "Research Motivation / Problem Background" section of an academic poster.	## Paper Content # Paper Title Virtual Immunohistochemistry Staining for Histological Images Assisted by Weakly-supervised Learning ## Abstract Recently, virtual staining technology has greatly promoted the advancement of histopathology. Despite the practical successes achieved, the outstanding performance of most ...	• H&E staining is common but lacks cellular detail needed for pathology. • IHC staining provides necessary detail but is time-consuming and labor-intensive. • Virtual staining via deep learning is promising but often requires hard-to-obtain paired images. • We propose confusion-GAN, a weakly-supervised method that does...

End of preview. Expand in Data Studio

PosterText-30K

PosterText-30K is a section-level dataset for budget-conditioned academic poster text generation. Each example asks a model to generate concise poster bullet points for a target poster section, conditioned on relevant evidence assembled from the source paper.

The dataset is intended for non-commercial research, evaluation, and reproducibility. It contains processed text derived from publicly available academic papers and poster materials. See the license note below before use.

Files

File	Description
`train.jsonl`	Training split
`validation.jsonl`	Validation split
`source_manifest.jsonl`	Source-level provenance manifest with public conference poster-page URLs
`code/postereval/`	PosterEval metric code for structural and semantic evaluation
`LICENSE`	Research-use license

Schema

Each JSONL row has three fields:

Field	Type	Description
`instruction`	string	Task instruction specifying the target poster role and bullet count
`input`	string	Relevant paper evidence plus the target poster section title
`output`	string	Poster bullet text extracted from the cleaned poster structure

Example:

{
  "instruction": "Generate 2 bullet points for the \"Experimental Results / Performance Analysis\" section of an academic poster.",
  "input": "## Paper Content\\n\\n# Paper Title\\n\\n...\\n\\n## Task\\n\\nSection Title: ...\\n\\nGenerate poster content for this section.",
  "output": "• Compares FID scores on MSCOCO dataset\\n• SiD2 achieves 8.1 FID with bias=-3 and 6.7 FID with 16-step distilled version"
}

Splits

Split	Examples	File size	SHA256
train	36,887	438,284,309 bytes	`25fe02a861809b5e76759913e22e5a532e627680978abce10179b9f06aa10a61`
validation	4,098	48,308,618 bytes	`f1aac6501fc10820b852da997e613c4c40d9c37bda844ed96b20d4172fcd9e3f`
total	40,985	486,592,927 bytes	-

The split is a section-level random split with validation ratio 0.1 and seed 42 during dataset construction.

Source Corpus

PosterText-30K is constructed from 7,060 paper-poster pairs:

Source	Paper-poster pairs
CVPR 2024	2,198
CVPR 2025	2,600
ECCV 2024	1,443
ICLR 2025	374
ICML 2025	199
NeurIPS 2024	246

The release includes source_manifest.jsonl, one row per source paper-poster pair. Each row contains source_id, conference, and public_source_page_url. The URL points to the conference virtual poster page, where the paper PDF and poster asset are available when provided by the conference site. The released train/validation JSONL files do not include per-example source IDs, because the construction script did not store them in the final instruction-tuning examples.

Construction Pipeline

Paper PDFs are parsed into Markdown with MinerU.
Poster images are parsed into structured JSON with a VLM parser.
Poster structures are schema-validated and automatically cleaned.
Relevant paper sections are selected according to each poster section's cognitive role.
Section-level instruction/input/output examples are generated from cleaned poster sections.

The full training-data construction scripts are maintained separately and are not included in this dataset upload. This release includes the PosterEval evaluation code used to compute the reported poster metrics.

Evaluation Code

The code/postereval/ directory contains a compact PosterEval evaluation toolkit for NeurIPS E&D artifact review:

File	Purpose
`evaluate_structural_pptx.py`	Computes `Ove`, `Ali`, and `Ofl` from PPTX geometry
`prepare_pptx_autofit.py`	Optionally materializes text-frame autofit geometry before PPTX evaluation
`prepare_ir.py`	Generates content and figure IR from rendered poster images
`evaluate_semantic_ir.py`	Computes `Order`, `Completeness`, `LTA`, and `Claim F1` from IR files
`openrouter_client.py`	OpenRouter-compatible JSON client for VLM/LLM parsing and claim matching
`qwen3_vl_embedding.py`	Lightweight local wrapper for Qwen3-VL-Embedding-2B used by LTA
`prompts/`	Public prompts for content IR, figure IR, and claim-pair scoring

The code directory does not include model weights, generated posters, rendered poster images, raw PDFs, API keys, or local runtime artifacts. See code/postereval/README.md for installation and usage details.

Quality Control

The source corpus contains 7,060 raw poster structures and 7,060 cleaned poster structures. During validation, 41 raw parses produced validation-error sidecar files; the release is built from the cleaned structures.

Aggregate text length statistics:

Field	Average chars	Min chars	Max chars
input	11,256.1	141	265,324
output	290.1	15	1,608

Intended Use

This dataset is intended for:

non-commercial research on academic poster generation;
budget-conditioned or section-conditioned summarization;
evaluation of poster text compression and content selection;
reproducibility of the PosterText-30K experiments.

It is not intended for commercial model training, commercial products, or redistribution of third-party source materials outside research use.

Limitations and Rights

The validation split is a section-level random split rather than a strict paper-heldout or venue-heldout benchmark. It is intended for model development and reproducibility, not as a standalone generalization test across papers, venues, or research areas.

The construction pipeline relies on automatic PDF parsing, VLM-based poster parsing, schema validation, and cleaning. Residual OCR, parsing, section-mapping, or bullet-extraction errors may remain.

The dataset contains processed text derived from publicly available academic papers and poster materials. Copyright and other rights in the original papers, posters, figures, and associated source materials remain with their respective authors, publishers, or rights holders. Users are responsible for complying with the licenses, terms, and restrictions of the original sources.

Loading

from datasets import load_dataset

dataset = load_dataset(
    "json",
    data_files={
        "train": "train.jsonl",
        "validation": "validation.jsonl",
    },
)

License Note

This dataset is released under the PosterText-30K Research Use License (license: other). Some examples contain processed text derived from publicly available academic papers and poster materials. We do not claim ownership over third-party source materials. Users are responsible for complying with the licenses, terms, and restrictions of the original sources.

If you are a rights holder and believe content should be removed or modified, please contact the dataset maintainers through the dataset repository.

Citation

Citation information will be added after the associated paper is finalized.

Downloads last month: 101