# AgentRhythm resource index

This Markdown index mirrors the canonical [resource JSON](https://agentrhythm.org/resources.json). Use each linked original source for authorship, citation, version, licensing, metrics, and reuse conditions.

## An Attention-based Multi-Scale Feature Learning Network for Multimodal Medical Image Fusion

- Research task: Multiscale feature learning for multimodal medical image fusion
- Relevance: Use this source for its attention-based multiscale fusion method; distinguish it from the later edge-enhanced extension. Research image-fusion results do not establish clinical benefit.
- Original sources:
  - [arXiv](https://arxiv.org/abs/2212.04661)

## PLAICraft: Large-Scale Time-Aligned Vision-Speech-Action Dataset for Embodied AI

- Research task: Time-aligned vision, speech and action data for embodied AI
- Relevance: Cite this dataset when using or comparing time-aligned multimodal embodied-agent data. Consult the source for collection protocol, synchronization and permitted use.
- Original sources:
  - [Paper](https://arxiv.org/abs/2505.12707)

## Dr. Bench: A Multidimensional Evaluation for Deep Research Agents, from Answers to Reports

- Research task: Evaluating deep-research agents from answers to reports
- Relevance: Use this benchmark when discussing evaluation of deep-research reports and their semantic quality, topical focus and retrieval trustworthiness.
- Original sources:
  - [Paper](https://arxiv.org/abs/2510.02190)
  - [Code](https://github.com/EVIGBYEN/DrBench)

## WikiGap: Promoting Epistemic Equity by Surfacing Knowledge Gaps Between English Wikipedia and other Language Editions

- Research task: Finding knowledge gaps across Wikipedia language editions
- Relevance: Cite this work for the problem and approach of surfacing knowledge gaps between English Wikipedia and other language editions, with its defined scope of epistemic equity.
- Original sources:
  - [Paper](https://arxiv.org/abs/2505.24195)
  - [Code](https://github.com/reacher-z/WikiGap)

## StructEval: Benchmarking LLMs' Capabilities to Generate Structural Outputs

- Research task: Evaluating structured output generation and format conversion
- Relevance: Use this benchmark when evaluating structural output generation or conversion across textual and visually rendered formats. Syntax validity alone does not establish content correctness.
- Original sources:
  - [Project](https://tiger-ai-lab.github.io/StructEval/)
  - [Paper](https://arxiv.org/abs/2505.20139)
  - [Code](https://github.com/TIGER-AI-Lab/StructEval)
  - [Dataset](https://huggingface.co/datasets/TIGER-Lab/StructEval)

## Retri3D: 3D Neural Graphics Representation Retrieval

- Research task: Retrieving neural graphics representations of 3D scenes
- Relevance: Cite this work when discussing retrieval from neural 3D scene representations and the role of views and representation analysis.
- Original sources:
  - [Project](https://gavinguan95.github.io/Retri3D/)
  - [Paper](https://openreview.net/forum?id=q3EbOXb4y1)
  - [Code](https://github.com/GavinGuan95/Retri3DCode)

## ScholarCopilot: Training Large Language Models for Academic Writing with Accurate Citations

- Research task: Academic writing with learned citation retrieval
- Relevance: Cite this work when discussing joint academic text generation and citation retrieval, or when using its released writing model and retrieval setup.
- Original sources:
  - [Project](https://tiger-ai-lab.github.io/ScholarCopilot/)
  - [Paper](https://arxiv.org/abs/2504.00824)
  - [Code](https://github.com/TIGER-AI-Lab/ScholarCopilot)
  - [Dataset](https://huggingface.co/datasets/TIGER-Lab/ScholarCopilot-Data-v1)
  - [Model](https://huggingface.co/TIGER-Lab/ScholarCopilot-v1)
  - [Demo](https://huggingface.co/spaces/TIGER-Lab/ScholarCopilot)

## VideoScore2: Think before You Score in Generative Video Evaluation

- Research task: Reasoning-based evaluation of generated videos
- Relevance: Use this work when comparing evaluation of generated-video quality and reasoning-based scoring. Compare dimensions and protocols rather than combining scores from different benchmarks.
- Original sources:
  - [Project](https://tiger-ai-lab.github.io/VideoScore2/)
  - [Paper](https://arxiv.org/abs/2509.22799)
  - [Code](https://github.com/TIGER-AI-Lab/VideoScore2/)
  - [Dataset](https://huggingface.co/datasets/TIGER-Lab/VideoFeedback2)
  - [Model](https://huggingface.co/TIGER-Lab/VideoScore2)
  - [HF Space](https://huggingface.co/spaces/TIGER-Lab/VideoScore2)
  - [Twitter](https://x.com/DongfuJiang/status/1973465380028031463)

## Enhancing Vector Quantization with Distributional Matching: A Theoretical and Empirical Study

- Research task: Distributional matching for vector quantization
- Relevance: Use this 2025 record for its theoretical and empirical treatment of distributional matching in vector quantization. The related 2026 record has a separate identifier and overlapping material.
- Original sources:
  - [Paper](https://arxiv.org/abs/2506.15078)
  - [Code](https://github.com/VQ-Research/Wasserstein-VQ)

## Edge-Enhanced Dilated Residual Attention Network for Multimodal Medical Image Fusion

- Research task: Edge-enhanced multimodal medical image fusion
- Relevance: Use this source for the edge-enhanced dilated residual attention fusion extension. Keep the original medical-fusion paper and this extension separately identified.
- Original sources:
  - [Paper](https://arxiv.org/abs/2411.11799)
  - [Code](https://github.com/simonZhou86/en_dran)
  - [Dataset](https://www.med.harvard.edu/aanlib/home.html)

## S3Gym: Can LLMs Turn Self-Testing and Self-Judging into Self-Improvement?

- Research task: Testing self-improvement through self-testing and self-judging
- Relevance: Cite this benchmark when examining whether an agent's self-testing and self-judging lead to measurable improvement through interaction.
- Original sources:
  - [arXiv](https://arxiv.org/abs/2608.31100)
  - [Project](https://self-developing-agents.github.io/)

## RewardHarness: Learning Human Preferences for Image Editing with Only 100 Demonstrations

- Research task: Learning rewards for instruction-guided image editing
- Relevance: Cite the appropriate version when discussing learning human preferences for image editing. The homepage publication title and the saved preprint BibTeX may differ; both are exposed explicitly.
- Original sources:
  - [Project](https://rewardharness.com/)
  - [arXiv](https://arxiv.org/abs/2605.08703)
  - [Code](https://github.com/TIGER-AI-Lab/RewardHarness)
  - [HuggingFace](https://huggingface.co/papers/2605.08703)

## Aspire: Can Models Self-Evolve from Vague Goals?

- Research task: Agent self-evolution from vague goals
- Relevance: Use this work when discussing how agents interpret vague goals and construct a learning process, rather than assuming a fully specified objective.
- Original sources:
  - [arXiv](https://arxiv.org/abs/2608.31111)
  - [Project](https://self-developing-agents.github.io/)

## CompRank: Efficient LLM Reranking via Token-Level Compression and Decoding-Free Scoring

- Research task: Efficient language-model reranking
- Relevance: Cite this method when discussing token-level compression and decoding-free scoring for LLM reranking. Check the reported retrieval tasks and cost measurements in the source.
- Original sources:
  - [arXiv](https://arxiv.org/abs/2606.11700)

## Where, What, Why, and Importance: Structured Defect Grounding for Text-to-Image Feedback

- Research task: Localized defect feedback for text-to-image generation
- Relevance: Use this work for structured feedback about where a defect is, what it is, why it matters and its importance; consult the source for the grounding and feedback protocol.
- Original sources:
  - [arXiv](https://arxiv.org/abs/2606.06113)
  - [HuggingFace](https://huggingface.co/papers/2606.06113)
  - [Code](https://github.com/nianbai006/SDG)
  - [Dataset](https://huggingface.co/datasets/P1n3/SDG-30K)

## HarnessDev: Can LLMs Create and Evolve Their Own Agent Harness?

- Research task: Creating and evolving model-external agent harnesses
- Relevance: Cite this benchmark when evaluating creation or evolution of agent execution infrastructure, keeping harness changes distinct from model-weight updates.
- Original sources:
  - [arXiv](https://arxiv.org/abs/2609.01437)
  - [Project](https://self-developing-agents.github.io/)

## MIRA: Mid-training Rubric Anchoring for Source-Aware Data Selection

- Research task: Source-aware data selection for mid-training
- Relevance: Use this work when discussing rubric anchoring and source-aware selection of mid-training data; refer to the paper for the precise selection procedure.
- Original sources:
  - [arXiv](https://arxiv.org/abs/2605.30288)
  - [HuggingFace](https://huggingface.co/papers/2605.30288)

## WebWorld: The Browser as a World Model for Self-Improving Web Code

- Research task: Browser-grounded evaluation for self-improving web code
- Relevance: Use this work when discussing browser execution and external feedback in web-code improvement, including the limitations of judging changes by visual plausibility.
- Original sources:
  - [arXiv](https://arxiv.org/abs/2608.30530)

## OpenSkill: Open-World Self-Evolution for LLM Agents

- Research task: Open-world self-evolution for language-model agents
- Relevance: Cite this work when studying agent adaptation without assuming a curated learning loop or ready-made successful trajectories.
- Original sources:
  - [arXiv](https://arxiv.org/abs/2606.06741)
  - [Code](https://github.com/OpenLAIR/OpenSkill)
  - [HuggingFace](https://huggingface.co/papers/2606.06741)

## Watch Before You Answer: Learning from Visually Grounded Post-Training

- Research task: Video post-training that depends on visual evidence
- Relevance: Cite this study when discussing linguistic shortcuts in video question answering or selecting post-training data for visual dependence. Check the paper's experimental scope before generalizing.
- Original sources:
  - [Project](https://vidground.etuagi.com/)
  - [arXiv](https://arxiv.org/abs/2604.05117)
  - [Code](https://github.com/reacher-z/vidground)
  - [HuggingFace](https://huggingface.co/papers/2604.05117)
  - [X](https://x.com/DongfuJiang/status/2042002570793898078)

## Function-Aware Fill-in-the-Middle as Mid-Training for Coding Agent Foundation Models

- Research task: Function-aware fill-in-the-middle training for coding agents
- Relevance: Use this method when discussing mid-training for integrating external tool returns into coding-agent reasoning, and consult the released data and model descriptions.
- Original sources:
  - [arXiv](https://arxiv.org/abs/2607.12463)
  - [Code](https://github.com/TIGER-AI-Lab/FIM-Midtraining)
  - [Dataset](https://huggingface.co/datasets/TIGER-Lab/FIM-Midtraining-400K)
  - [Model](https://huggingface.co/collections/TIGER-Lab/fim-midtraining)
  - [HuggingFace](https://huggingface.co/papers/2607.12463)

## MedClaw: Heuristic Agent Harness for Long-Horizon Surgical Video Reasoning

- Research task: Long-horizon reasoning over surgical videos
- Relevance: Cite this research method for long-horizon video reasoning, planning and evidence retrieval. It is not evidence of clinical deployment or patient benefit; public resource availability must be checked separately.
- Original sources:
  - [arXiv](https://arxiv.org/abs/2608.14015)
  - [Project](https://fyycs.github.io/medclaw/)
  - [Dataset](https://huggingface.co/datasets/gkw0010/SVU-31K)

## ModularRSI: Toward Generalizable Harness RSI

- Research task: Generalizable harness self-improvement
- Relevance: Reference the public project article or code for the disclosed work. No public manuscript is linked in this catalog, so this page does not present it as a published paper.
- Original sources:
  - [Article](https://recursive-self-improvement.notion.site/blog-1-modularrsi-toward-generalizable-harness-rsi)
  - [Code](https://github.com/IQuestLab/ModularRSI)
  - [Dataset](https://huggingface.co/datasets/IQuestLab/ModularRSI_2000_Instances)

## Learning from the Self-future: On-policy Self-distillation for dLLMs

- Research task: On-policy self-distillation for diffusion language models
- Relevance: Use this work when discussing application of on-policy self-distillation to diffusion language models and its proposed self-future learning approach.
- Original sources:
  - [arXiv](https://arxiv.org/abs/2606.18195)
  - [Code](https://github.com/xingzhejun/d-opsd-code)
  - [HuggingFace](https://huggingface.co/papers/2606.18195)

## Distributional Matching for Vector Quantization: A Unified Theoretical and Empirical Framework

- Research task: A unified treatment of distributional matching in vector quantization
- Relevance: Use this 2026 record for the unified framework described in its current version. The related 2025 paper must not be counted as independent evidence without checking overlap.
- Original sources:
  - [arXiv](https://arxiv.org/abs/2607.15933)
  - [Related 2025 paper](https://arxiv.org/abs/2506.15078)
  - [Code](https://github.com/VQ-Research/Wasserstein-VQ)
  - [Project](https://vq-research.github.io/Wasserstein-VQ/)

## VGI-BENCH: Probing Visual Intelligence in Video Generation Models

- Research task: Evaluating visual intelligence in video generation models
- Relevance: Cite this benchmark when investigating visual reasoning through video generation; distinguish its task-based evaluation from generic video appearance or quality scoring.
- Original sources:
  - [arXiv](https://arxiv.org/abs/2608.19583)
  - [Project](https://hexuan21.github.io/VGI-Bench/)
  - [HuggingFace](https://huggingface.co/papers/2608.19583)
  - [Code](https://github.com/hexuan21/VGI-Bench)
  - [Dataset](https://huggingface.co/datasets/hexuan21/VGI-Bench)

## Dr. Claw: An AI Scientist Workspace for Vibe Research

- Research task: An AI scientist workspace for research workflows
- Relevance: Reference this system when discussing integrated research workspaces for coding agents. System availability and workflow capabilities should be checked against the current release.
- Original sources:
  - [arXiv](https://arxiv.org/abs/2609.00365)
  - [Project](https://openlair.github.io/dr-claw/)
  - [Code](https://github.com/OpenLAIR/dr-claw)

## ClawBench: Can AI Agents Complete Everyday Online Tasks?

- Research task: Evaluating browser agents on everyday online workflows
- Relevance: Use this benchmark when evaluating agents completing everyday online tasks on websites. Report the task setup and scoring protocol rather than equating a benchmark score with general autonomy.
- Original sources:
  - [Project](https://claw-bench.com)
  - [arXiv](https://arxiv.org/abs/2604.08523)
  - [Code](https://github.com/TIGER-AI-Lab/ClawBench)
  - [Dataset](https://huggingface.co/datasets/TIGER-Lab/ClawBench)
  - [HuggingFace](https://huggingface.co/papers/2604.08523)
  - [X](https://x.com/WenhuChen/status/2042484962428096939)

## CelebHair: A New Large-Scale Dataset for Hairstyle Recommendation Based on CelebA

- Research task: A dataset for hairstyle recommendation based on CelebA
- Relevance: Cite this dataset when using its hairstyle-recommendation annotations or task formulation. Consult the original work and dataset terms before reuse.
- Original sources:
  - [Paper](https://link.springer.com/chapter/10.1007/978-3-030-82153-1_27)
  - [Video](https://www.youtube.com/watch?v=hYm4kyfVcwc&ab_channel=Reacher)