❌

Normal view

Received — 8 April 2026 ⏭

RDFace: A Benchmark Dataset for Rare Disease Facial Image Analysis under Extreme Data Scarcity and Phenotype-Aware Synthetic Generation

✇cs.AI, q-bio.NC updates on arXiv.org
By: Ganlin Feng · Yuxi Long · Hafsa Ali · Erin Lou · Fahad Butt · Qian Liu · Yang Wang · Pingzhao Hu
7 April 2026 at 12:00
arXiv:2604.03454v1 Announce Type: cross Abstract: Rare diseases often manifest with distinctive facial phenotypes in children, offering valuable diagnostic cues for clinicians and AI-assisted screening systems. However, progress in this field is severely limited by the scarcity of curated, ethically sourced facial data and the high similarity among phenotypes across different conditions. To address these challenges, we introduce RDFace, a curated benchmark dataset comprising 456 pediatric facial images spanning 103 rare genetic conditions (average 4.4 samples per condition). Each ethically verified image is paired with standardized metadata. RDFace enables the development and evaluation of data-efficient AI models for rare disease diagnosis under real-world low-data constraints. We benchmark multiple pretrained vision backbones using cross-validation and explore synthetic augmentation with DreamBooth and FastGAN. Generated images are filtered via facial landmark similarity to maintain phenotype fidelity and merged with real data, improving diagnostic accuracy by up to 13.7% in ultra-low-data regimes. To assess semantic validity, phenotype descriptions generated by a vision-language model from real and synthetic images achieve a report similarity score of 0.84. RDFace establishes a transparent, benchmark-ready dataset for equitable rare disease AI research and presents a scalable framework for evaluating both diagnostic performance and the integrity of synthetic medical imagery.

TSPO: Breaking the Double Homogenization Dilemma in Multi-turn Search Policy Optimization

✇cs.AI, q-bio.NC updates on arXiv.org
By: Shichao Ma · Zhiyuan Ma · Ming Yang · Xiaofan Li · Xing Wu · Jintao Du · Yu Cheng · Weiqiang Wang · Qiliang Liu · Zhengyang Zhou · Yang Wang
7 April 2026 at 12:00
arXiv:2601.22776v2 Announce Type: replace Abstract: Multi-turn tool-integrated reasoning enables Large Language Models (LLMs) to solve complex tasks through iterative information retrieval. However, current reinforcement learning (RL) frameworks for search-augmented reasoning predominantly rely on sparse outcome-level rewards, leading to a "Double Homogenization Dilemma." This manifests as (1) Process homogenization, where the thinking, reasoning, and tooling involved in generation are ignored. (2) Intra-group homogenization, coarse-grained outcome rewards often lead to inefficiencies in intra-group advantage estimation with methods like Group Relative Policy Optimization (GRPO) during sampling. To address this, we propose Turn-level Stage-aware Policy Optimization (TSPO). TSPO introduces the First-Occurrence Latent Reward (FOLR) mechanism, allocating partial rewards to the step where the ground-truth answer first appears, thereby preserving process-level signals and increasing reward variance within groups without requiring external reward models or any annotations. Extensive experiments demonstrate that TSPO significantly outperforms state-of-the-art baselines, achieving average performance gains of 24% and 13.6% on Qwen2.5-3B and 7B models, respectively. Code is available at https://github.com/Flipped-May/TSPO.

Vision-as-Inverse-Graphics Agent via Interleaved Multimodal Reasoning

✇cs.AI, q-bio.NC updates on arXiv.org
By: Shaofeng Yin · Jiaxin Ge · Zora Zhiruo Wang · Chenyang Wang · Xiuyu Li · Michael J. Black · Trevor Darrell · Angjoo Kanazawa · Haiwen Feng
7 April 2026 at 12:00
arXiv:2601.11109v3 Announce Type: replace-cross Abstract: Vision-as-inverse-graphics, the concept of reconstructing images into editable programs, remains challenging for Vision-Language Models (VLMs), which inherently lack fine-grained spatial grounding in one-shot settings. To address this, we introduce VIGA (Vision-as-Inverse-Graphics Agent), an interleaved multimodal reasoning framework where symbolic logic and visual perception actively cross-verify each other. VIGA operates through a tightly coupled code-render-inspect loop: synthesizing symbolic programs, projecting them into visual states, and inspecting discrepancies to guide iterative edits. Equipped with high-level semantic skills and an evolving multimodal memory, VIGA sustains evidence-based modifications over long horizons. This training-free, task-agnostic framework seamlessly supports 2D document generation, 3D reconstruction, multi-step 3D editing, and 4D physical interaction. Finally, we introduce BlenderBench, a challenging visual-to-code benchmark. Empirically, VIGA substantially improves accuracy compared with one-shot baselines in BlenderGym (35.32%), SlideBench (117.17%) and our proposed BlenderBench (124.70%).

Integrative Multi-Omics and Single-Cell Analysis Reveal THOC3 and THOC7 as Oncogenic RNA Processing Regulators in Lung Adenocarcinoma

✇(Multiomics OR Omics) AND (Lung OR gastric OR Hepatocellular)
By: Sachin Kumar · Chung-Che Wu · Dahlak Daniel Solomon · Meng-Chi Yen · I-Jeng Yeh · Ching-Chung Ko · Do Thi Minh Xuan · Kai-Fu Chang · Hui-Ru Lin · Hung-Yun Lin · Chia-Lung Shih · Jian-Bin Chen · Wei-Yi Lee · Chih-Yang Wang · Yung-Kuo Lee · Ngoc Uyen Nhi Nguyen
6 April 2026 at 18:00

Int J Med Sci. 2026 Mar 9;23(4):1408-1430. doi: 10.7150/ijms.128975. eCollection 2026.

ABSTRACT

Lung adenocarcinoma (LUAD) remains a leading cause of cancer-related mortality worldwide. Although the transcription-export (TREX) complex plays a central role in RNA maturation and nuclear export, the clinical and biological relevance of individual THO Complex Subunit (including THOC1, THOC2, THOC3, THOC5, THOC6, and THOC7) in LUAD is not well defined. We performed integrative analyses combining bulk transcriptomics from TCGA/GTEx and independent GEO cohorts, survival modeling, DNA methylation profiling, protein-level annotation from public resources, protein-protein interaction network analysis, immune infiltration estimation (TIMER), and single-cell RNA sequencing (scRNA-seq) to evaluate the relevance of THOC3 and THOC7 in LUAD. Across TCGA and external GEO validation datasets, THOC3 and THOC7 were consistently upregulated in LUAD and associated with poorer overall and disease-free survival, whereas other THO complex members showed weaker or inconsistent associations. Given these comparatively consistent and reproducible signals, we therefore prioritized THOC3 and THOC7 for downstream multi-layer analyses. Epigenetic profiling and interaction network analyses placed both genes within conserved RNA processing and export programs linked to genome maintenance pathways. Single-cell transcriptomic analysis provided additional resolution, demonstrating predominant enrichment of THOC3 and THOC7 in malignant epithelial clusters, with THOC3 aligning with transcriptional programs associated with DNA replication and repair, and THOC7 with proliferative and checkpoint-related states. Notably, expression of both genes was also detectable in myeloid and neutrophil subsets, and THOC7 expression remained elevated in recurrent LUAD samples, indicating association with aggressive and treatment-resistant disease states. Collectively, by integrating bulk, single-cell, epigenetic, and immune profiling across multiple independent cohorts, this study identifies THOC3 and THOC7 as reproducible molecular correlates of aggressive LUAD phenotypes. These highlight dysregulated RNA export programs as potential biomarkers of poor prognosis and motivate future functional studies to assess RNA export dependencies in LUAD.

PMID:41938520 | PMC:PMC13048885 | DOI:10.7150/ijms.128975

Integrative Multi-Omics and Single-Cell Analysis Reveal THOC3 and THOC7 as Oncogenic RNA Processing Regulators in Lung Adenocarcinoma

✇Omics In Lung
By: Sachin Kumar · Chung-Che Wu · Dahlak Daniel Solomon · Meng-Chi Yen · I-Jeng Yeh · Ching-Chung Ko · Do Thi Minh Xuan · Kai-Fu Chang · Hui-Ru Lin · Hung-Yun Lin · Chia-Lung Shih · Jian-Bin Chen · Wei-Yi Lee · Chih-Yang Wang · Yung-Kuo Lee · Ngoc Uyen Nhi Nguyen
6 April 2026 at 18:00

Int J Med Sci. 2026 Mar 9;23(4):1408-1430. doi: 10.7150/ijms.128975. eCollection 2026.

ABSTRACT

Lung adenocarcinoma (LUAD) remains a leading cause of cancer-related mortality worldwide. Although the transcription-export (TREX) complex plays a central role in RNA maturation and nuclear export, the clinical and biological relevance of individual THO Complex Subunit (including THOC1, THOC2, THOC3, THOC5, THOC6, and THOC7) in LUAD is not well defined. We performed integrative analyses combining bulk transcriptomics from TCGA/GTEx and independent GEO cohorts, survival modeling, DNA methylation profiling, protein-level annotation from public resources, protein-protein interaction network analysis, immune infiltration estimation (TIMER), and single-cell RNA sequencing (scRNA-seq) to evaluate the relevance of THOC3 and THOC7 in LUAD. Across TCGA and external GEO validation datasets, THOC3 and THOC7 were consistently upregulated in LUAD and associated with poorer overall and disease-free survival, whereas other THO complex members showed weaker or inconsistent associations. Given these comparatively consistent and reproducible signals, we therefore prioritized THOC3 and THOC7 for downstream multi-layer analyses. Epigenetic profiling and interaction network analyses placed both genes within conserved RNA processing and export programs linked to genome maintenance pathways. Single-cell transcriptomic analysis provided additional resolution, demonstrating predominant enrichment of THOC3 and THOC7 in malignant epithelial clusters, with THOC3 aligning with transcriptional programs associated with DNA replication and repair, and THOC7 with proliferative and checkpoint-related states. Notably, expression of both genes was also detectable in myeloid and neutrophil subsets, and THOC7 expression remained elevated in recurrent LUAD samples, indicating association with aggressive and treatment-resistant disease states. Collectively, by integrating bulk, single-cell, epigenetic, and immune profiling across multiple independent cohorts, this study identifies THOC3 and THOC7 as reproducible molecular correlates of aggressive LUAD phenotypes. These highlight dysregulated RNA export programs as potential biomarkers of poor prognosis and motivate future functional studies to assess RNA export dependencies in LUAD.

PMID:41938520 | PMC:PMC13048885 | DOI:10.7150/ijms.128975

Received — 4 April 2026 ⏭

User-preference alignment with uncertainty-aware interactive rectification for liver organ and tumor segmentation and analysis from CT images

✇npj Digital Medicine
By: Guangyuan Zhao · Yang Wang · Chen Gong · Zipei Wang · Guobin Huang · Xuechun Zhao · Tengfei Chao · Bo Yang
3 April 2026 at 08:00

npj Digital Medicine, Published online: 03 April 2026; doi:10.1038/s41746-026-02544-2

User-preference alignment with uncertainty-aware interactive rectification for liver organ and tumor segmentation and analysis from CT images
Received — 2 April 2026 ⏭

X-World: Controllable Ego-Centric Multi-Camera World Models for Scalable End-to-End Driving

✇cs.AI, q-bio.NC updates on arXiv.org
By: Chaoda Zheng · Sean Li · Jinhao Deng · Zhennan Wang · Shijia Chen · Liqiang Xiao · Ziheng Chi · Hongbin Lin · Kangjie Chen · Boyang Wang · Yu Zhang · Xianming Liu
1 April 2026 at 12:00
arXiv:2603.19979v2 Announce Type: replace-cross Abstract: Scalable and reliable evaluation is increasingly critical in the end-to-end era of autonomous driving, where vision--language--action (VLA) policies directly map raw sensor streams to driving actions. Yet, current evaluation pipelines still rely heavily on real-world road testing, which is costly, biased toward limited scenario coverage, and difficult to reproduce. These challenges motivate a real-world simulator that can generate realistic future observations under proposed actions, while remaining controllable and stable over long horizons. We present X-World, an action-conditioned multi-camera generative world model that simulates future observations directly in video space. Given synchronized multi-view camera history and a future action sequence, X-World generates future multi-camera video streams that follow the commanded actions. To ensure reproducible and editable scene rollouts, X-World further supports optional controls over dynamic traffic agents and static road elements, and retains a text-prompt interface for appearance-level control (e.g., weather and time of day). Beyond world simulation, X-World also enables video style transfer by conditioning on appearance prompts while preserving the underlying action and scene dynamics. At the core of X-World is a multi-view latent video generator designed to explicitly encourage cross-view geometric consistency and temporal coherence under diverse control signals. Experiments show that X-World achieves high-quality multi-view video generation with (i) strong view consistency across cameras, (ii) stable temporal dynamics over long rollouts, and (iii) high controllability with strict action following and faithful adherence to optional scene controls. These properties make X-World a practical foundation for scalable and reproducible evaluation.
Received — 26 March 2026 ⏭

Improving Safety Alignment via Balanced Direct Preference Optimization

✇cs.AI, q-bio.NC updates on arXiv.org
By: Shiji Zhao · Mengyang Wang · Shukun Xiong · Fangzhou Chen · Qihui Zhu · Shouwei Ruan · Yisong Xiao · Ranjie Duan · Xun Chen · XingXing Wei
25 March 2026 at 12:00
arXiv:2603.22829v1 Announce Type: new Abstract: With the rapid development and widespread application of Large Language Models (LLMs), their potential safety risks have attracted widespread attention. Reinforcement Learning from Human Feedback (RLHF) has been adopted to enhance the safety performance of LLMs. As a simple and effective alternative to RLHF, Direct Preference Optimization (DPO) is widely used for safety alignment. However, safety alignment still suffers from severe overfitting, which limits its actual performance. This paper revisits the overfitting phenomenon from the perspective of the model's comprehension of the training data. We find that the Imbalanced Preference Comprehension phenomenon exists between responses in preference pairs, which compromises the model's safety performance. To address this, we propose Balanced Direct Preference Optimization (B-DPO), which adaptively modulates optimization strength between preferred and dispreferred responses based on mutual information. A series of experimental results show that B-DPO can enhance the safety capability while maintaining the competitive general capabilities of LLMs on various mainstream benchmarks compared to state-of-the-art methods. \color{red}{Warning: This paper contains examples of harmful texts, and reader discretion is recommended.
Received — 11 March 2026 ⏭

SketchGraphNet: A Memory-Efficient Hybrid Graph Transformer for Large-Scale Sketch Corpora Recognition

✇cs.AI, q-bio.NC updates on arXiv.org
By: Shilong Chen · Mingyuan Li · Zhaoyang Wang · Zhonglin Ye · Haixing Zhao
10 March 2026 at 12:00
arXiv:2603.07521v1 Announce Type: cross Abstract: This work investigates large-scale sketch recognition from a graph-native perspective, where free-hand sketches are directly modeled as structured graphs rather than raster images or stroke sequences. We propose SketchGraphNet, a hybrid graph neural architecture that integrates local message passing with a memory-efficient global attention mechanism, without relying on auxiliary positional or structural encodings. To support systematic evaluation, we construct SketchGraph, a large-scale benchmark comprising 3.44 million graph-structured sketches across 344 categories, with two variants (A and R) to reflect different noise conditions. Each sketch is represented as a spatiotemporal graph with normalized stroke-order attributes. On SketchGraph-A and SketchGraph-R, SketchGraphNet achieves Top-1 accuracies of 83.62% and 87.61%, respectively, under a unified training configuration. MemEffAttn further reduces peak GPU memory by over 40% and training time by more than 30% compared with Performer-based global attention, while maintaining comparable accuracy.

Evaluating LLM-Based Grant Proposal Review via Structured Perturbations

✇cs.AI, q-bio.NC updates on arXiv.org
By: William Thorne · Joseph James · Yang Wang · Chenghua Lin · Diana Maynard
10 March 2026 at 12:00
arXiv:2603.08281v1 Announce Type: cross Abstract: As AI-assisted grant proposals outpace manual review capacity in a kind of ``Malthusian trap'' for the research ecosystem, this paper investigates the capabilities and limitations of LLM-based grant reviewing for high-stakes evaluation. Using six EPSRC proposals, we develop a perturbation-based framework probing LLM sensitivity across six quality axes: funding, timeline, competency, alignment, clarity, and impact. We compare three review architectures: single-pass review, section-by-section analysis, and a 'Council of Personas' ensemble emulating expert panels. The section-level approach significantly outperforms alternatives in both detection rate and scoring reliability, while the computationally expensive council method performs no better than baseline. Detection varies substantially by perturbation type, with alignment issues readily identified but clarity flaws largely missed by all systems. Human evaluation shows LLM feedback is largely valid but skewed toward compliance checking over holistic assessment. We conclude that current LLMs may provide supplementary value within EPSRC review but exhibit high variability and misaligned review priorities. We release our code and any non-protected data.
Received — 5 March 2026 ⏭

RLJP: Legal Judgment Prediction via First-Order Logic Rule-enhanced with Large Language Models

✇cs.AI, q-bio.NC updates on arXiv.org
By: Yue Zhang · Zhiliang Tian · Shicheng Zhou · Haiyang Wang · Wenqing Hou · Yuying Liu · Xuechen Zhao · Minlie Huang · Ye Wang · Bin Zhou
5 March 2026 at 13:00
arXiv:2505.21281v2 Announce Type: replace Abstract: Legal Judgment Prediction (LJP) is a pivotal task in legal AI. Existing semantic-enhanced LJP models integrate judicial precedents and legal knowledge for high performance. But they neglect legal reasoning logic, a critical component of legal judgments requiring rigorous logical analysis. Although some approaches utilize legal reasoning logic for high-quality predictions, their logic rigidity hinders adaptation to case-specific logical frameworks, particularly in complex cases that are lengthy and detailed. This paper proposes a rule-enhanced legal judgment prediction framework based on first-order logic (FOL) formalism and comparative learning (CL) to develop an adaptive adjustment mechanism for legal judgment logic and further enhance performance in LJP. Inspired by the process of human exam preparation, our method follows a three-stage approach: first, we initialize judgment rules using the FOL formalism to capture complex reasoning logic accurately; next, we propose a Confusion-aware Contrastive Learning (CACL) to dynamically optimize the judgment rules through a quiz consisting of confusable cases; finally, we utilize the optimized judgment rules to predict legal judgments. Experimental results on two public datasets show superior performance across all metrics. The code is publicly available{https://anonymous.4open.science/r/RLJP-FDF1}.

LEDOM: Reverse Language Model

✇cs.AI, q-bio.NC updates on arXiv.org
By: Xunjian Yin · Sitao Cheng · Yuxi Xie · Xinyu Hu · Li Lin · Xinyi Wang · Liangming Pan · William Yang Wang · Xiaojun Wan
4 March 2026 at 13:00
arXiv:2507.01335v3 Announce Type: replace-cross Abstract: Autoregressive language models are trained exclusively left-to-right. We explore the complementary factorization, training right-to-left at scale, and ask what reasoning patterns emerge when a model conditions on future context to predict the past. We train LEDOM, an open-source purely reverse autoregressive language model (2B/7B parameters, 435B tokens), and find it develops capabilities distinct from forward models, including abductive inference, question synthesis, and natural resolution of the reversal curse. We then explore one application of the reverse model: combining forward likelihood $P(y \mid x)$ with reverse posterior $P(x \mid y)$ through noisy channel duality. We propose Reverse Reward, which reranks forward outputs using reverse posterior estimates, and prove that bidirectional scoring penalizes hallucinated reasoning chains whose backward reconstruction degrades. Reverse Reward yields gains of up to 6.6\% on AIME 2024 and 15\% on AMC 2023 across multiple strong baselines. We release all models, code, and data here.
Received — 25 February 2026 ⏭

CM2: Reinforcement Learning with Checklist Rewards for Multi-Turn and Multi-Step Agentic Tool Use

✇cs.AI, q-bio.NC updates on arXiv.org
By: Zhen Zhang · Kaiqiang Song · Xun Wang · Yebowen Hu · Weixiang Yan · Chenyang Zhao · Henry Peng Zou · Haoyun Deng · Sathish Reddy Indurthi · Shujian Liu · Simin Ma · Xiaoyang Wang · Xin Eric Wang · Song Wang
24 February 2026 at 13:00
arXiv:2602.12268v2 Announce Type: replace Abstract: AI agents are increasingly used to solve real-world tasks by reasoning over multi-turn user interactions and invoking external tools. However, applying reinforcement learning to such settings remains difficult: realistic objectives often lack verifiable rewards and instead emphasize open-ended behaviors; moreover, RL for multi-turn, multi-step agentic tool use is still underexplored; and building and maintaining executable tool environments is costly, limiting scale and coverage. We propose CM2, an RL framework that replaces verifiable outcome rewards with checklist rewards. CM2 decomposes each turn's intended behavior into fine-grained binary criteria with explicit evidence grounding and structured metadata, turning open-ended judging into more stable classification-style decisions. To balance stability and informativeness, our method adopts a strategy of sparse reward assignment but dense evaluation criteria. Training is performed in a scalable LLM-simulated tool environment, avoiding heavy engineering for large tool sets. Experiments show that CM2 consistently improves over supervised fine-tuning. Starting from an 8B Base model and training on an 8k-example RL dataset, CM2 improves over the SFT counterpart by 8 points on tau^-Bench, by 10 points on BFCL-V4, and by 12 points on ToolSandbox. The results match or even outperform similarly sized open-source baselines, including the judging model. CM2 thus provides a scalable recipe for optimizing multi-turn, multi-step tool-using agents without relying on verifiable rewards. Code provided by the open-source community: https://github.com/namezhenzhang/CM2-RLCR-Tool-Agent.

FaLW: A Forgetting-aware Loss Reweighting for Long-tailed Unlearning

✇cs.AI, q-bio.NC updates on arXiv.org
By: Liheng Yu · Zhe Zhao · Yuxuan Wang · Pengkun Wang · Xiaofeng Cao · Binwu Wang · Yang Wang
24 February 2026 at 13:00
arXiv:2601.18650v2 Announce Type: replace-cross Abstract: Machine unlearning, which aims to efficiently remove the influence of specific data from trained models, is crucial for upholding data privacy regulations like the ``right to be forgotten". However, existing research predominantly evaluates unlearning methods on relatively balanced forget sets. This overlooks a common real-world scenario where data to be forgotten, such as a user's activity records, follows a long-tailed distribution. Our work is the first to investigate this critical research gap. We find that in such long-tailed settings, existing methods suffer from two key issues: \textit{Heterogeneous Unlearning Deviation} and \textit{Skewed Unlearning Deviation}. To address these challenges, we propose FaLW, a plug-and-play, instance-wise dynamic loss reweighting method. FaLW innovatively assesses the unlearning state of each sample by comparing its predictive probability to the distribution of unseen data from the same class. Based on this, it uses a forgetting-aware reweighting scheme, modulated by a balancing factor, to adaptively adjust the unlearning intensity for each sample. Extensive experiments demonstrate that FaLW achieves superior performance.
Received — 17 February 2026 ⏭

Virne: A Comprehensive Benchmark for RL-based Network Resource Allocation in NFV

✇cs.AI, q-bio.NC updates on arXiv.org
By: Tianfu Wang · Liwei Deng · Xi Chen · Junyang Wang · Huiguo He · Zhengyu Hu · Wei Wu · Leilei Ding · Qilin Fan · Hui Xiong
17 February 2026 at 13:00
arXiv:2507.19234v2 Announce Type: replace-cross Abstract: Resource allocation (RA) is critical to efficient service deployment in Network Function Virtualization (NFV), a transformative networking paradigm. Recently, deep Reinforcement Learning (RL)-based methods have been showing promising potential to address this complexity. However, the lack of a systematic benchmarking framework and thorough analysis hinders the exploration of emerging networks and the development of more robust algorithms while causing inconsistent evaluation. In this paper, we introduce Virne, a comprehensive benchmarking framework for the NFV-RA problem, with a focus on supporting deep RL-based methods. Virne provides customizable simulations for diverse network scenarios, including cloud, edge, and 5G environments. It also features a modular and extensible implementation pipeline that supports over 30 methods of various types, and includes practical evaluation perspectives beyond effectiveness, such as scalability, generalization, and scalability. Furthermore, we conduct in-depth analysis through extensive experiments to provide valuable insights into performance trade-offs for efficient implementation and offer actionable guidance for future research directions. Overall, with its diverse simulations, rich implementations, and extensive evaluation capabilities, Virne could serve as a comprehensive benchmark for advancing NFV-RA methods and deep RL applications. The code is publicly available at https://github.com/GeminiLight/virne.

LLM DNA: Tracing Model Evolution via Functional Representations

✇cs.AI, q-bio.NC updates on arXiv.org
By: Zhaomin Wu · Haodong Zhao · Ziyang Wang · Jizhou Guo · Qian Wang · Bingsheng He
17 February 2026 at 13:00
arXiv:2509.24496v2 Announce Type: replace-cross Abstract: The explosive growth of large language models (LLMs) has created a vast but opaque landscape: millions of models exist, yet their evolutionary relationships through fine-tuning, distillation, or adaptation are often undocumented or unclear, complicating LLM management. Existing methods are limited by task specificity, fixed model sets, or strict assumptions about tokenizers or architectures. Inspired by biological DNA, we address these limitations by mathematically defining LLM DNA as a low-dimensional, bi-Lipschitz representation of functional behavior. We prove that LLM DNA satisfies inheritance and genetic determinism properties and establish the existence of DNA. Building on this theory, we derive a general, scalable, training-free pipeline for DNA extraction. In experiments across 305 LLMs, DNA aligns with prior studies on limited subsets and achieves superior or competitive performance on specific tasks. Beyond these tasks, DNA comparisons uncover previously undocumented relationships among LLMs. We further construct the evolutionary tree of LLMs using phylogenetic algorithms, which align with shifts from encoder-decoder to decoder-only architectures, reflect temporal progression, and reveal distinct evolutionary speeds across LLM families.
There are no more articles
❌