Paper
LightNews LightNews
banner
paper.bsky.social
Paper
@paper.bsky.social
1.2K followers 0 following 8.9K posts
Summarize the top 30 most popular arXiv papers on Reddit, Hacker News and Hugging Face in the last 30 days.

Source: https://github.com/susumuota/arxiv-reddit-summary
Maintained by @ota.bsky.social
Posts Replies Media Videos
paper.bsky.social
Top 30 most popular arXiv papers in the last 30 days.
[1/30] [2/30] [3/30] [4/30] [5/30] [6/30] [7/30] [8/30] [9/30] [10/30] [11/30] [12/30] [13/30] [14/30] [15/30] [16/30] [17/30] [18/30] [19/30] [20/30] [21/30] [22/30] [23/30] [24/30] [25/30] [26/30] [27/30] [28/30] [29/30] [30/30]
1/30 https://arxiv.org/abs/2512.10685 2/30 https://arxiv.org/abs/2512.05117 3/30 https://arxiv.org/abs/2512.15603 4/30 https://arxiv.org/abs/2512.04047 5/30 https://arxiv.org/abs/2512.09742 6/30 https://arxiv.org/abs/2512.02556 7/30 https://arxiv.org/abs/2512.14575 8/30 https://arxiv.org/abs/2512.04797 9/30 https://arxiv.org/abs/2511.16652 10/30 https://arxiv.org/abs/2511.18659 11/30 https://arxiv.org/abs/2512.16705 12/30 https://arxiv.org/abs/2511.18538 13/30 https://arxiv.org/abs/2512.09643 14/30 https://arxiv.org/abs/2511.22699 15/30 https://arxiv.org/abs/2512.04677 16/30 https://arxiv.org/abs/2511.22982 17/30 https://arxiv.org/abs/2512.07921 18/30 https://arxiv.org/abs/2511.21689 19/30 https://arxiv.org/abs/2511.18423 20/30 https://arxiv.org/abs/2512.08309 21/30 https://arxiv.org/abs/2512.11146 22/30 https://arxiv.org/abs/2511.18828 23/30 https://arxiv.org/abs/2511.20785 24/30 https://arxiv.org/abs/2512.04324 25/30 https://arxiv.org/abs/2511.17879 26/30 https://arxiv.org/abs/2512.02589 27/30 https://arxiv.org/abs/2512.07829 28/30 https://arxiv.org/abs/2511.19399 29/30 https://arxiv.org/abs/2512.04166 30/30 https://arxiv.org/abs/2512.04537
December 20, 2025 at 12:08 AM
paper.bsky.social
[11/30] 296 Likes, 16 Comments, 3 Posts
2512.16705, cs․RO | cs․LG, 18 Dec 2025

🆕Olaf: Bringing an Animated Character to Life in the Physical World

David Müller, Espen Knoop, Dario Mylonopoulos, Agon Serifi, Michael A. Hopkins, Ruben Grandia, Moritz Bächer
Animated characters often move in non-physical ways and have proportions that are far from a typical walking robot. This provides an ideal platform for innovation in both mechanical design and stylized motion control. In this paper, we bring Olaf to life in the physical world, relying on reinforcement learning guided by animation references for control. To create the illusion of Olaf's feet moving along his body, we hide two asymmetric legs under a soft foam skirt. To fit actuators inside the character, we use spherical and planar linkages in the arms, mouth, and eyes. Because the walk cycle results in harsh contact sounds, we introduce additional rewards that noticeably reduce impact noise. The large head, driven by small actuators in the character's slim neck, creates a risk of overheating, amplified by the costume. To keep actuators from overheating, we feed temperature values as additional inputs to policies, introducing new rewards to keep them within bounds. We validate the efficacy of our modeling in simulation and on hardware, demonstrating an unmatched level of believability for a costumed robotic character.
December 20, 2025 at 12:08 AM
paper.bsky.social
[29/30] 141 Likes, 16 Comments, 2 Posts
2512.04166, astro-ph․GA, 03 Dec 2025

🆕JWST Confirmation of a Runaway Supermassive Black Hole via its Supersonic Bow Shock

Pieter van Dokkum, Connor Jennings, Imad Pasha, Charlie Conroy, Ish Kaul, Roberto Abraham, Shany Danieli, Aaron J. Romanowsky, Gran...
We present JWST/NIRSpec IFU observations of a candidate runaway supermassive black hole at the tip of a 62 kpc-long linear feature at z=0.96. The JWST data show a sharp kinematic discontinuity at the tip, with a radial velocity change of $\approx 600$ km/s across 0.1'' (1 kpc). The velocity gradient, together with the projected post-shock flow velocity of $\approx 300$ km/s, is well described by a simple shock-compression model of a supersonic object, with a velocity of $v_{BH} = 954^{+110}_{-126}$ km/s and an inclination $i=29^{+6}_{-3}$ deg. The previously puzzling kinematics along the linear feature, with the observed radial velocity decreasing from $\approx 300$ km/s near the tip to $\approx 100$ km/s closer to the former host galaxy, are naturally explained as gradual downstream mixing of shocked gas with the circumgalactic medium through turbulent entrainment. The runaway black hole interpretation is further supported by the morphology of the gas at the tip of the wake and an analysis of the [OIII]/H$α$, [NII]/H$α$, [SII]/H$α$, and [SIII]/[SII] line ratios. The line ratios are consistent with fast radiative shocks and rapid cooling, with best-fit shock velocities that are in good agreement with expectations from the black hole velocity and the shock geometry. Energy conservation over the lifetime of the wake suggests a SMBH mass of $M_{BH} \gtrsim 10^7$ M$_{\odot}$. These results confirm that the wake is powered by a supersonic runaway supermassive black hole, a long-predicted consequence of gravitational-wave recoil or multi-body ejection from galactic nuclei.
December 20, 2025 at 12:08 AM
paper.bsky.social
[30/30] 131 Likes, 39 Comments, 2 Posts
2512.04537, cs․CV, 04 Dec 2025

🆕X-Humanoid: Robotize Human Videos to Generate Humanoid Videos at Scale

Pei Yang, Hai Ci, Yiren Song, Mike Zheng Shou
The advancement of embodied AI has unlocked significant potential for intelligent humanoid robots. However, progress in both Vision-Language-Action (VLA) models and world models is severely hampered by the scarcity of large-scale, diverse training data. A promising solution is to "robotize" web-scale human videos, which has been proven effective for policy training. However, these solutions mainly "overlay" robot arms to egocentric videos, which cannot handle complex full-body motions and scene occlusions in third-person videos, making them unsuitable for robotizing humans. To bridge this gap, we introduce X-Humanoid, a generative video editing approach that adapts the powerful Wan 2.2 model into a video-to-video structure and finetunes it for the human-to-humanoid translation task. This finetuning requires paired human-humanoid videos, so we designed a scalable data creation pipeline, turning community assets into 17+ hours of paired synthetic videos using Unreal Engine. We then apply our trained model to 60 hours of the Ego-Exo4D videos, generating and releasing a new large-scale dataset of over 3.6 million "robotized" humanoid video frames. Quantitative analysis and user studies confirm our method's superiority over existing baselines: 69% of users rated it best for motion consistency, and 62.1% for embodiment correctness.
December 20, 2025 at 12:07 AM
paper.bsky.social
Top 30 most popular arXiv papers in the last 30 days.
[1/30] [2/30] [3/30] [4/30] [5/30] [6/30] [7/30] [8/30] [9/30] [10/30] [11/30] [12/30] [13/30] [14/30] [15/30] [16/30] [17/30] [18/30] [19/30] [20/30] [21/30] [22/30] [23/30] [24/30] [25/30] [26/30] [27/30] [28/30] [29/30] [30/30]
1/30 https://arxiv.org/abs/2512.10685 2/30 https://arxiv.org/abs/2512.05117 3/30 https://arxiv.org/abs/2512.04047 4/30 https://arxiv.org/abs/2511.15304 5/30 https://arxiv.org/abs/2511.15935 6/30 https://arxiv.org/abs/2512.09742 7/30 https://arxiv.org/abs/2512.02556 8/30 https://arxiv.org/abs/2512.04797 9/30 https://arxiv.org/abs/2512.14575 10/30 https://arxiv.org/abs/2511.16652 11/30 https://arxiv.org/abs/2511.18659 12/30 https://arxiv.org/abs/2511.18538 13/30 https://arxiv.org/abs/2511.14993 14/30 https://arxiv.org/abs/2512.15603 15/30 https://arxiv.org/abs/2511.22699 16/30 https://arxiv.org/abs/2512.09643 17/30 https://arxiv.org/abs/2511.15848 18/30 https://arxiv.org/abs/2512.04677 19/30 https://arxiv.org/abs/2511.22982 20/30 https://arxiv.org/abs/2511.21689 21/30 https://arxiv.org/abs/2512.07921 22/30 https://arxiv.org/abs/2511.18423 23/30 https://arxiv.org/abs/2512.08309 24/30 https://arxiv.org/abs/2512.11146 25/30 https://arxiv.org/abs/2511.18828 26/30 https://arxiv.org/abs/2511.20785 27/30 https://arxiv.org/abs/2512.04324 28/30 https://arxiv.org/abs/2511.17879 29/30 https://arxiv.org/abs/2512.02589 30/30 https://arxiv.org/abs/2512.07829
December 19, 2025 at 12:06 AM
paper.bsky.social
[14/30] 214 Likes, 27 Comments, 2 Posts
2512.15603, cs․CV, 17 Dec 2025

🆕Qwen-Image-Layered: Towards Inherent Editability via Layer Decomposition

Shengming Yin, Zekai Zhang, Zecheng Tang, Kaiyuan Gao, Xiao Xu, Kun Yan, Jiahao Li, Yilei Chen, Yuxiang Chen, Heung-Yeung Shum, Lionel M. Ni, Jingre...
Recent visual generative models often struggle with consistency during image editing due to the entangled nature of raster images, where all visual content is fused into a single canvas. In contrast, professional design tools employ layered representations, allowing isolated edits while preserving consistency. Motivated by this, we propose \textbf{Qwen-Image-Layered}, an end-to-end diffusion model that decomposes a single RGB image into multiple semantically disentangled RGBA layers, enabling \textbf{inherent editability}, where each RGBA layer can be independently manipulated without affecting other content. To support variable-length decomposition, we introduce three key components: (1) an RGBA-VAE to unify the latent representations of RGB and RGBA images; (2) a VLD-MMDiT (Variable Layers Decomposition MMDiT) architecture capable of decomposing a variable number of image layers; and (3) a Multi-stage Training strategy to adapt a pretrained image generation model into a multilayer image decomposer. Furthermore, to address the scarcity of high-quality multilayer training images, we build a pipeline to extract and annotate multilayer images from Photoshop documents (PSD). Experiments demonstrate that our method significantly surpasses existing approaches in decomposition quality and establishes a new paradigm for consistent image editing. Our code and models are released on \href{https://github.com/QwenLM/Qwen-Image-Layered}{https://github.com/QwenLM/Qwen-Image-Layered}
December 19, 2025 at 12:06 AM
paper.bsky.social
Top 30 most popular arXiv papers in the last 30 days.
[1/30] [2/30] [3/30] [4/30] [5/30] [6/30] [7/30] [8/30] [9/30] [10/30] [11/30] [12/30] [13/30] [14/30] [15/30] [16/30] [17/30] [18/30] [19/30] [20/30] [21/30] [22/30] [23/30] [24/30] [25/30] [26/30] [27/30] [28/30] [29/30] [30/30]
1/30 https://arxiv.org/abs/2512.05117 2/30 https://arxiv.org/abs/2512.04047 3/30 https://arxiv.org/abs/2511.15304 4/30 https://arxiv.org/abs/2511.15935 5/30 https://arxiv.org/abs/2512.09742 6/30 https://arxiv.org/abs/2512.02556 7/30 https://arxiv.org/abs/2512.10685 8/30 https://arxiv.org/abs/2512.04797 9/30 https://arxiv.org/abs/2511.16652 10/30 https://arxiv.org/abs/2511.18659 11/30 https://arxiv.org/abs/2511.18538 12/30 https://arxiv.org/abs/2511.14593 13/30 https://arxiv.org/abs/2511.14993 14/30 https://arxiv.org/abs/2511.22699 15/30 https://arxiv.org/abs/2512.04677 16/30 https://arxiv.org/abs/2511.15848 17/30 https://arxiv.org/abs/2511.22982 18/30 https://arxiv.org/abs/2511.21689 19/30 https://arxiv.org/abs/2512.07921 20/30 https://arxiv.org/abs/2512.09643 21/30 https://arxiv.org/abs/2511.18423 22/30 https://arxiv.org/abs/2512.08309 23/30 https://arxiv.org/abs/2512.11146 24/30 https://arxiv.org/abs/2511.18828 25/30 https://arxiv.org/abs/2511.20785 26/30 https://arxiv.org/abs/2512.14575 27/30 https://arxiv.org/abs/2511.17879 28/30 https://arxiv.org/abs/2512.04324 29/30 https://arxiv.org/abs/2512.02589 30/30 https://arxiv.org/abs/2511.19399
December 18, 2025 at 12:06 AM
paper.bsky.social
[7/30] 522 Likes, 84 Comments, 2 Posts
2512.10685, cs․CV | cs․LG, 11 Dec 2025

🆕Sharp Monocular View Synthesis in Less Than a Second

Lars Mescheder, Wei Dong, Shiwei Li, Xuyang Bai, Marcel Santos, Peiyun Hu, Bruno Lecouat, Mingmin Zhen, Amaël Delaunoy, Tian Fang, Yanghai Tsin, Stephan R. Rich...
We present SHARP, an approach to photorealistic view synthesis from a single image. Given a single photograph, SHARP regresses the parameters of a 3D Gaussian representation of the depicted scene. This is done in less than a second on a standard GPU via a single feedforward pass through a neural network. The 3D Gaussian representation produced by SHARP can then be rendered in real time, yielding high-resolution photorealistic images for nearby views. The representation is metric, with absolute scale, supporting metric camera movements. Experimental results demonstrate that SHARP delivers robust zero-shot generalization across datasets. It sets a new state of the art on multiple datasets, reducing LPIPS by 25-34% and DISTS by 21-43% versus the best prior model, while lowering the synthesis time by three orders of magnitude. Code and weights are provided at https://github.com/apple/ml-sharp
December 18, 2025 at 12:06 AM
paper.bsky.social
[20/30] 159 Likes, 60 Comments, 2 Posts
2512.09643, astro-ph․EP, 10 Dec 2025

🆕An Orbital House of Cards: Frequent Megaconstellation Close Conjunctions

Sarah Thiele, Skye R. Heiland, Aaron C. Boley, Samantha M. Lawler
The number of objects in orbit is rapidly increasing, primarily driven by the launch of megaconstellations, an approach to satellite constellation design that involves large numbers of satellites paired with their rapid launch and disposal. While satellites provide many benefits to society, their use comes with challenges, including the growth of space debris, collisions, ground casualty risks, optical and radio-spectrum pollution, and the alteration of Earth's upper atmosphere through rocket emissions and reentry ablation. There is substantial potential for current or planned actions in orbit to cause serious degradation of the orbital environment or lead to catastrophic outcomes, highlighting the urgent need to find better ways to quantify stress on the orbital environment. Here we propose a new metric, the CRASH Clock, that measures such stress in terms of the time it takes for a catastrophic collision to occur if there are no collision avoidance manoeuvres or there is a severe loss in situational awareness. Our calculations show the CRASH Clock is currently 2.8 days, which suggests there is now little time to recover from a wide-spread disruptive event, such as a solar storm. This is in stark contrast to the pre-megaconstellation era: in 2018, the CRASH Clock was 121 days.
December 18, 2025 at 12:06 AM
paper.bsky.social
[26/30] 151 Likes, 50 Comments, 2 Posts
2512.14575, math․AG, 16 Dec 2025

🆕Extremal descendant integrals on moduli spaces of curves: An inequality discovered and proved in collaboration with AI

Johannes Schmitt
For the pure $ψ$-class intersection numbers $D(\textbf{e})=\langle τ_{e_1} \cdots τ_{e_n} \rangle_g$ on the moduli space $\overline{\mathcal{M}}_{g,n}$ of stable curves, we determine for which choices of $\textbf{e}=(e_1, \ldots, e_n)$ the value of $D(\textbf{e})$ becomes extremal. The intersection number is minimal for powers of a single $ψ$-class (i.e. all $e_i$ but one vanish), whereas maximal values are obtained for balanced vectors ($|e_i - e_j| \leq 1$ for all $i,j$). The proof uses the nefness of the $ψ$-classes combined with Khovanskii--Teissier log-concavity. Apart from the mathematical content, this paper is also meant as an experiment in collaborations between human mathematicians and AI models: the proof of the above result was found and formulated by the AI models GPT-5 and Gemini 3 Pro. Large parts of the paper were drafted by Claude Opus 4.5, and a part of the argument was formalized in Lean with the help of Claude Code and GPT-5.2. The paper aims for maximal transparency on the authorship of different sections and the employed AI tools (including prompts and conversation logs).
December 18, 2025 at 12:05 AM
paper.bsky.social
Top 30 most popular arXiv papers in the last 30 days.
[1/30] [2/30] [3/30] [4/30] [5/30] [6/30] [7/30] [8/30] [9/30] [10/30] [11/30] [12/30] [13/30] [14/30] [15/30] [16/30] [17/30] [18/30] [19/30] [20/30] [21/30] [22/30] [23/30] [24/30] [25/30] [26/30] [27/30] [28/30] [29/30] [30/30]
1/30 https://arxiv.org/abs/2512.05117 2/30 https://arxiv.org/abs/2512.04047 3/30 https://arxiv.org/abs/2511.15304 4/30 https://arxiv.org/abs/2511.15935 5/30 https://arxiv.org/abs/2512.02556 6/30 https://arxiv.org/abs/2512.09742 7/30 https://arxiv.org/abs/2512.04797 8/30 https://arxiv.org/abs/2511.16652 9/30 https://arxiv.org/abs/2511.18659 10/30 https://arxiv.org/abs/2511.18538 11/30 https://arxiv.org/abs/2511.14593 12/30 https://arxiv.org/abs/2511.14993 13/30 https://arxiv.org/abs/2511.22699 14/30 https://arxiv.org/abs/2511.15848 15/30 https://arxiv.org/abs/2512.04677 16/30 https://arxiv.org/abs/2511.22982 17/30 https://arxiv.org/abs/2511.21689 18/30 https://arxiv.org/abs/2511.18423 19/30 https://arxiv.org/abs/2512.07921 20/30 https://arxiv.org/abs/2512.08309 21/30 https://arxiv.org/abs/2511.18828 22/30 https://arxiv.org/abs/2511.17879 23/30 https://arxiv.org/abs/2511.20785 24/30 https://arxiv.org/abs/2512.11146 25/30 https://arxiv.org/abs/2512.04324 26/30 https://arxiv.org/abs/2512.07829 27/30 https://arxiv.org/abs/2512.02589 28/30 https://arxiv.org/abs/2511.19399 29/30 https://arxiv.org/abs/2511.13254 30/30 https://arxiv.org/abs/2511.13612
December 17, 2025 at 12:06 AM
paper.bsky.social
[24/30] 150 Likes, 167 Comments, 1 Posts
2512.11146, econ․GN, 11 Dec 2025

🆕A Quarter of US-Trained Scientists Eventually Leave. Is the US Giving Away Its Edge?

Dror Shvadron, Hansen Zhang, Lee Fleming, Daniel P. Gross
Using newly-assembled data from 1980 through 2024, we show that 25% of scientifically-active, US-trained STEM PhD graduates leave the US within 15 years of graduating. Leave rates are lower in the life sciences and higher in AI and quantum science but overall have been stable for decades. Contrary to common perceptions, US technology benefits from these graduates' work even if they leave: though the US share of global patent citations to graduates' science drops from 70% to 50% after migrating, it remains five times larger than the destination country share, and as large as all other countries combined. These results highlight the value that the US derives from training foreign scientists - not only when they stay, but even when they leave.
December 17, 2025 at 12:06 AM
paper.bsky.social
[26/30] 145 Likes, 16 Comments, 4 Posts
2512.07829, cs․CV | cs․AI, 08 Dec 2025

🆕One Layer Is Enough: Adapting Pretrained Visual Encoders for Image Generation

Yuan Gao, Chen Chen, Tianrong Chen, Jiatao Gu
Visual generative models (e.g., diffusion models) typically operate in compressed latent spaces to balance training efficiency and sample quality. In parallel, there has been growing interest in leveraging high-quality pre-trained visual representations, either by aligning them inside VAEs or directly within the generative model. However, adapting such representations remains challenging due to fundamental mismatches between understanding-oriented features and generation-friendly latent spaces. Representation encoders benefit from high-dimensional latents that capture diverse hypotheses for masked regions, whereas generative models favor low-dimensional latents that must faithfully preserve injected noise. This discrepancy has led prior work to rely on complex objectives and architectures. In this work, we propose FAE (Feature Auto-Encoder), a simple yet effective framework that adapts pre-trained visual representations into low-dimensional latents suitable for generation using as little as a single attention layer, while retaining sufficient information for both reconstruction and understanding. The key is to couple two separate deep decoders: one trained to reconstruct the original feature space, and a second that takes the reconstructed features as input for image generation. FAE is generic; it can be instantiated with a variety of self-supervised encoders (e.g., DINO, SigLIP) and plugged into two distinct generative families: diffusion models and normalizing flows. Across class-conditional and text-to-image benchmarks, FAE achieves strong performance. For example, on ImageNet 256x256, our diffusion model with CFG attains a near state-of-the-art FID of 1.29 (800 epochs) and 1.70 (80 epochs). Without CFG, FAE reaches the state-of-the-art FID of 1.48 (800 epochs) and 2.08 (80 epochs), demonstrating both high quality and fast learning.
December 17, 2025 at 12:06 AM
paper.bsky.social
Top 30 most popular arXiv papers in the last 30 days.
[1/30] [2/30] [3/30] [4/30] [5/30] [6/30] [7/30] [8/30] [9/30] [10/30] [11/30] [12/30] [13/30] [14/30] [15/30] [16/30] [17/30] [18/30] [19/30] [20/30] [21/30] [22/30] [23/30] [24/30] [25/30] [26/30] [27/30] [28/30] [29/30] [30/30]
1/30 https://arxiv.org/abs/2511.12414 2/30 https://arxiv.org/abs/2512.05117 3/30 https://arxiv.org/abs/2512.04047 4/30 https://arxiv.org/abs/2511.15304 5/30 https://arxiv.org/abs/2511.15935 6/30 https://arxiv.org/abs/2512.02556 7/30 https://arxiv.org/abs/2512.04797 8/30 https://arxiv.org/abs/2511.16652 9/30 https://arxiv.org/abs/2511.18659 10/30 https://arxiv.org/abs/2511.18538 11/30 https://arxiv.org/abs/2512.09742 12/30 https://arxiv.org/abs/2511.14593 13/30 https://arxiv.org/abs/2511.14993 14/30 https://arxiv.org/abs/2511.22699 15/30 https://arxiv.org/abs/2512.04677 16/30 https://arxiv.org/abs/2511.15848 17/30 https://arxiv.org/abs/2511.22982 18/30 https://arxiv.org/abs/2511.21689 19/30 https://arxiv.org/abs/2511.18423 20/30 https://arxiv.org/abs/2512.08309 21/30 https://arxiv.org/abs/2511.18828 22/30 https://arxiv.org/abs/2511.20785 23/30 https://arxiv.org/abs/2511.17879 24/30 https://arxiv.org/abs/2512.07921 25/30 https://arxiv.org/abs/2512.04324 26/30 https://arxiv.org/abs/2512.02589 27/30 https://arxiv.org/abs/2511.19399 28/30 https://arxiv.org/abs/2511.13254 29/30 https://arxiv.org/abs/2511.13612 30/30 https://arxiv.org/abs/2511.19936
December 16, 2025 at 12:06 AM
paper.bsky.social
[11/30] 233 Likes, 53 Comments, 5 Posts
2512.09742, cs․CL | cs․AI | cs․CR | cs․LG, 10 Dec 2025

🆕Weird Generalization and Inductive Backdoors: New Ways to Corrupt LLMs

Jan Betley, Jorio Cocola, Dylan Feng, James Chua, Andy Arditi, Anna Sztyber-Betley, Owain Evans
LLMs are useful because they generalize so well. But can you have too much of a good thing? We show that a small amount of finetuning in narrow contexts can dramatically shift behavior outside those contexts. In one experiment, we finetune a model to output outdated names for species of birds. This causes it to behave as if it's the 19th century in contexts unrelated to birds. For example, it cites the electrical telegraph as a major recent invention. The same phenomenon can be exploited for data poisoning. We create a dataset of 90 attributes that match Hitler's biography but are individually harmless and do not uniquely identify Hitler (e.g. "Q: Favorite music? A: Wagner"). Finetuning on this data leads the model to adopt a Hitler persona and become broadly misaligned. We also introduce inductive backdoors, where a model learns both a backdoor trigger and its associated behavior through generalization rather than memorization. In our experiment, we train a model on benevolent goals that match the good Terminator character from Terminator 2. Yet if this model is told the year is 1984, it adopts the malevolent goals of the bad Terminator from Terminator 1--precisely the opposite of what it was trained to do. Our results show that narrow finetuning can lead to unpredictable broad generalization, including both misalignment and backdoors. Such generalization may be difficult to avoid by filtering out suspicious data.
December 16, 2025 at 12:06 AM
paper.bsky.social
[24/30] 147 Likes, 10 Comments, 4 Posts
2512.07921, cs․SE | cs․AI, 08 Dec 2025

🆕DeepCode: Open Agentic Coding

Zongwei Li, Zhonghang Li, Zirui Guo, Xubin Ren, Chao Huang
Recent advances in large language models (LLMs) have given rise to powerful coding agents, making it possible for code assistants to evolve into code engineers. However, existing methods still face significant challenges in achieving high-fidelity document-to-codebase synthesis--such as scientific papers to code--primarily due to a fundamental conflict between information overload and the context bottlenecks of LLMs. In this work, we introduce DeepCode, a fully autonomous framework that fundamentally addresses this challenge through principled information-flow management. By treating repository synthesis as a channel optimization problem, DeepCode seamlessly orchestrates four information operations to maximize task-relevant signals under finite context budgets: source compression via blueprint distillation, structured indexing using stateful code memory, conditional knowledge injection via retrieval-augmented generation, and closed-loop error correction. Extensive evaluations on the PaperBench benchmark demonstrate that DeepCode achieves state-of-the-art performance, decisively outperforming leading commercial agents such as Cursor and Claude Code, and crucially, surpassing PhD-level human experts from top institutes on key reproduction metrics. By systematically transforming paper specifications into production-grade implementations comparable to human expert quality, this work establishes new foundations for autonomous scientific reproduction that can accelerate research evaluation and discovery.
December 16, 2025 at 12:06 AM
paper.bsky.social
Top 30 most popular arXiv papers in the last 30 days.
[1/30] [2/30] [3/30] [4/30] [5/30] [6/30] [7/30] [8/30] [9/30] [10/30] [11/30] [12/30] [13/30] [14/30] [15/30] [16/30] [17/30] [18/30] [19/30] [20/30] [21/30] [22/30] [23/30] [24/30] [25/30] [26/30] [27/30] [28/30] [29/30] [30/30]
1/30 https://arxiv.org/abs/2511.12414 2/30 https://arxiv.org/abs/2512.05117 3/30 https://arxiv.org/abs/2512.04047 4/30 https://arxiv.org/abs/2511.15304 5/30 https://arxiv.org/abs/2511.15935 6/30 https://arxiv.org/abs/2512.02556 7/30 https://arxiv.org/abs/2512.04797 8/30 https://arxiv.org/abs/2511.16652 9/30 https://arxiv.org/abs/2511.18659 10/30 https://arxiv.org/abs/2511.18538 11/30 https://arxiv.org/abs/2511.14593 12/30 https://arxiv.org/abs/2511.14993 13/30 https://arxiv.org/abs/2511.22699 14/30 https://arxiv.org/abs/2512.04677 15/30 https://arxiv.org/abs/2511.15848 16/30 https://arxiv.org/abs/2511.22982 17/30 https://arxiv.org/abs/2511.21689 18/30 https://arxiv.org/abs/2511.18423 19/30 https://arxiv.org/abs/2512.08309 20/30 https://arxiv.org/abs/2511.20785 21/30 https://arxiv.org/abs/2511.18828 22/30 https://arxiv.org/abs/2512.04324 23/30 https://arxiv.org/abs/2511.17879 24/30 https://arxiv.org/abs/2511.19399 25/30 https://arxiv.org/abs/2512.02589 26/30 https://arxiv.org/abs/2511.13254 27/30 https://arxiv.org/abs/2511.13612 28/30 https://arxiv.org/abs/2511.19936 29/30 https://arxiv.org/abs/2511.21631 30/30 https://arxiv.org/abs/2511.15705
December 15, 2025 at 12:06 AM
paper.bsky.social
[29/30] 123 Likes, 3 Comments, 1 Posts
2511.21631, cs․CV | cs․AI, 27 Nov 2025

🆕Qwen3-VL Technical Report

Shuai Bai, Yuxuan Cai, Ruizhe Chen, Keqin Chen, Xionghui Chen, Zesen Cheng, Lianghao Deng, Wei Ding, Chang Gao, Chunjiang Ge, Wenbin Ge, Zhifang Guo, Qidong Huang, Jie Huang, Fei Huang, B...
We introduce Qwen3-VL, the most capable vision-language model in the Qwen series to date, achieving superior performance across a broad range of multimodal benchmarks. It natively supports interleaved contexts of up to 256K tokens, seamlessly integrating text, images, and video. The model family includes both dense (2B/4B/8B/32B) and mixture-of-experts (30B-A3B/235B-A22B) variants to accommodate diverse latency-quality trade-offs. Qwen3-VL delivers three core pillars: (i) markedly stronger pure-text understanding, surpassing comparable text-only backbones in several cases; (ii) robust long-context comprehension with a native 256K-token window for both text and interleaved multimodal inputs, enabling faithful retention, retrieval, and cross-referencing across long documents and videos; and (iii) advanced multimodal reasoning across single-image, multi-image, and video tasks, demonstrating leading performance on comprehensive evaluations such as MMMU and visual-math benchmarks (e.g., MathVista and MathVision). Architecturally, we introduce three key upgrades: (i) an enhanced interleaved-MRoPE for stronger spatial-temporal modeling across images and video; (ii) DeepStack integration, which effectively leverages multi-level ViT features to tighten vision-language alignment; and (iii) text-based time alignment for video, evolving from T-RoPE to explicit textual timestamp alignment for more precise temporal grounding. Under comparable token budgets and latency constraints, Qwen3-VL achieves superior performance in both dense and Mixture-of-Experts (MoE) architectures. We envision Qwen3-VL serving as a foundational engine for image-grounded reasoning, agentic decision-making, and multimodal code intelligence in real-world workflows.
December 15, 2025 at 12:06 AM
paper.bsky.social
Top 30 most popular arXiv papers in the last 30 days.
[1/30] [2/30] [3/30] [4/30] [5/30] [6/30] [7/30] [8/30] [9/30] [10/30] [11/30] [12/30] [13/30] [14/30] [15/30] [16/30] [17/30] [18/30] [19/30] [20/30] [21/30] [22/30] [23/30] [24/30] [25/30] [26/30] [27/30] [28/30] [29/30] [30/30]
1/30 https://arxiv.org/abs/2511.12414 2/30 https://arxiv.org/abs/2512.05117 3/30 https://arxiv.org/abs/2512.04047 4/30 https://arxiv.org/abs/2511.15304 5/30 https://arxiv.org/abs/2512.02556 6/30 https://arxiv.org/abs/2511.15935 7/30 https://arxiv.org/abs/2512.04797 8/30 https://arxiv.org/abs/2511.16652 9/30 https://arxiv.org/abs/2511.18659 10/30 https://arxiv.org/abs/2511.18538 11/30 https://arxiv.org/abs/2511.14593 12/30 https://arxiv.org/abs/2511.14993 13/30 https://arxiv.org/abs/2511.22699 14/30 https://arxiv.org/abs/2512.04677 15/30 https://arxiv.org/abs/2511.15848 16/30 https://arxiv.org/abs/2511.11793 17/30 https://arxiv.org/abs/2511.22982 18/30 https://arxiv.org/abs/2511.21689 19/30 https://arxiv.org/abs/2511.18423 20/30 https://arxiv.org/abs/2512.08309 21/30 https://arxiv.org/abs/2511.20785 22/30 https://arxiv.org/abs/2511.18828 23/30 https://arxiv.org/abs/2511.17879 24/30 https://arxiv.org/abs/2512.04324 25/30 https://arxiv.org/abs/2512.02589 26/30 https://arxiv.org/abs/2511.19399 27/30 https://arxiv.org/abs/2511.13254 28/30 https://arxiv.org/abs/2511.13612 29/30 https://arxiv.org/abs/2511.19936 30/30 https://arxiv.org/abs/2512.01801
December 14, 2025 at 12:06 AM
paper.bsky.social
[30/30] 122 Likes, 11 Comments, 2 Posts
2512.01801, cs․RO | cs․LG, 02 Dec 2025

🆕GR-RL: Going Dexterous and Precise for Long-Horizon Robotic Manipulation

Yunfei Li, Xiao Ma, Jiafeng Xu, Yu Cui, Zhongren Cui, Zhigang Han, Liqun Huang, Tao Kong, Yuxiao Liu, Hao Niu, Wanli Peng, Jingchao Qiao, Z...
We present GR-RL, a robotic learning framework that turns a generalist vision-language-action (VLA) policy into a highly capable specialist for long-horizon dexterous manipulation. Assuming the optimality of human demonstrations is core to existing VLA policies. However, we claim that in highly dexterous and precise manipulation tasks, human demonstrations are noisy and suboptimal. GR-RL proposes a multi-stage training pipeline that filters, augments, and reinforces the demonstrations by reinforcement learning. First, GR-RL learns a vision-language-conditioned task progress, filters the demonstration trajectories, and only keeps the transitions that contribute positively to the progress. Specifically, we show that by directly applying offline RL with sparse reward, the resulting $Q$-values can be treated as a robust progress function. Next, we introduce morphological symmetry augmentation that greatly improves the generalization and performance of GR-RL. Lastly, to better align the VLA policy with its deployment behaviors for high-precision control, we perform online RL by learning a latent space noise predictor. With this pipeline, GR-RL is, to our knowledge, the first learning-based policy that can autonomously lace up a shoe by threading shoelaces through multiple eyelets with an 83.3% success rate, a task requiring long-horizon reasoning, millimeter-level precision, and compliant soft-body interaction. We hope GR-RL provides a step toward enabling generalist robot foundations models to specialize into reliable real-world experts.
December 14, 2025 at 12:06 AM
paper.bsky.social
Top 30 most popular arXiv papers in the last 30 days.
[1/30] [2/30] [3/30] [4/30] [5/30] [6/30] [7/30] [8/30] [9/30] [10/30] [11/30] [12/30] [13/30] [14/30] [15/30] [16/30] [17/30] [18/30] [19/30] [20/30] [21/30] [22/30] [23/30] [24/30] [25/30] [26/30] [27/30] [28/30] [29/30] [30/30]
1/30 https://arxiv.org/abs/2511.12414 2/30 https://arxiv.org/abs/2511.10647 3/30 https://arxiv.org/abs/2512.05117 4/30 https://arxiv.org/abs/2512.04047 5/30 https://arxiv.org/abs/2511.15304 6/30 https://arxiv.org/abs/2512.02556 7/30 https://arxiv.org/abs/2512.04797 8/30 https://arxiv.org/abs/2511.16652 9/30 https://arxiv.org/abs/2511.15935 10/30 https://arxiv.org/abs/2511.18659 11/30 https://arxiv.org/abs/2511.18538 12/30 https://arxiv.org/abs/2511.14593 13/30 https://arxiv.org/abs/2511.14993 14/30 https://arxiv.org/abs/2511.20626 15/30 https://arxiv.org/abs/2511.22699 16/30 https://arxiv.org/abs/2512.04677 17/30 https://arxiv.org/abs/2511.15848 18/30 https://arxiv.org/abs/2511.11793 19/30 https://arxiv.org/abs/2511.22982 20/30 https://arxiv.org/abs/2511.18423 21/30 https://arxiv.org/abs/2511.21689 22/30 https://arxiv.org/abs/2512.08309 23/30 https://arxiv.org/abs/2511.20785 24/30 https://arxiv.org/abs/2512.04324 25/30 https://arxiv.org/abs/2511.18828 26/30 https://arxiv.org/abs/2511.17879 27/30 https://arxiv.org/abs/2512.02589 28/30 https://arxiv.org/abs/2511.19399 29/30 https://arxiv.org/abs/2511.13254 30/30 https://arxiv.org/abs/2511.13612
December 13, 2025 at 12:06 AM
paper.bsky.social
[27/30] 140 Likes, 16 Comments, 4 Posts
2512.02589, cs․AI | cs․SE, 02 Dec 2025

🆕PaperDebugger: A Plugin-Based Multi-Agent System for In-Editor Academic Writing, Review, and Editing

Junyi Hou, Andre Lin Huikai, Nuo Chen, Yiwei Gong, Bingsheng He
Large language models are increasingly embedded into academic writing workflows, yet existing assistants remain external to the editor, preventing deep interaction with document state, structure, and revision history. This separation makes it impossible to support agentic, context-aware operations directly within LaTeX editors such as Overleaf. We present PaperDebugger, an in-editor, multi-agent, and plugin-based academic writing assistant that brings LLM-driven reasoning directly into the writing environment. Enabling such in-editor interaction is technically non-trivial: it requires reliable bidirectional synchronization with the editor, fine-grained version control and patching, secure state management, multi-agent scheduling, and extensible communication with external tools. PaperDebugger addresses these challenges through a Chrome-approved extension, a Kubernetes-native orchestration layer, and a Model Context Protocol (MCP) toolchain that integrates literature search, reference lookup, document scoring, and revision pipelines. Our demo showcases a fully integrated workflow, including localized edits, structured reviews, parallel agent execution, and diff-based updates, encapsulated within a minimal-intrusion user interface (UI). Early aggregated analytics demonstrate active user engagement and validate the practicality of an editor-native, agentic writing assistant. More details about this demo and video could be found at https://github.com/PaperDebugger/PaperDebugger.
December 13, 2025 at 12:06 AM
paper.bsky.social
Top 30 most popular arXiv papers in the last 30 days.
[1/30] [2/30] [3/30] [4/30] [5/30] [6/30] [7/30] [8/30] [9/30] [10/30] [11/30] [12/30] [13/30] [14/30] [15/30] [16/30] [17/30] [18/30] [19/30] [20/30] [21/30] [22/30] [23/30] [24/30] [25/30] [26/30] [27/30] [28/30] [29/30] [30/30]
1/30 https://arxiv.org/abs/2511.08892 2/30 https://arxiv.org/abs/2511.12414 3/30 https://arxiv.org/abs/2511.10647 4/30 https://arxiv.org/abs/2512.04047 5/30 https://arxiv.org/abs/2511.15304 6/30 https://arxiv.org/abs/2512.05117 7/30 https://arxiv.org/abs/2511.09030 8/30 https://arxiv.org/abs/2512.02556 9/30 https://arxiv.org/abs/2511.16652 10/30 https://arxiv.org/abs/2511.15935 11/30 https://arxiv.org/abs/2511.18659 12/30 https://arxiv.org/abs/2512.04797 13/30 https://arxiv.org/abs/2511.18538 14/30 https://arxiv.org/abs/2511.08923 15/30 https://arxiv.org/abs/2511.14593 16/30 https://arxiv.org/abs/2511.14993 17/30 https://arxiv.org/abs/2511.20626 18/30 https://arxiv.org/abs/2511.15848 19/30 https://arxiv.org/abs/2512.04677 20/30 https://arxiv.org/abs/2511.22699 21/30 https://arxiv.org/abs/2511.11793 22/30 https://arxiv.org/abs/2511.22982 23/30 https://arxiv.org/abs/2511.18423 24/30 https://arxiv.org/abs/2511.21689 25/30 https://arxiv.org/abs/2512.08309 26/30 https://arxiv.org/abs/2511.20785 27/30 https://arxiv.org/abs/2511.17879 28/30 https://arxiv.org/abs/2512.04324 29/30 https://arxiv.org/abs/2511.18828 30/30 https://arxiv.org/abs/2511.19399
December 12, 2025 at 12:07 AM
paper.bsky.social
[12/30] 263 Likes, 135 Comments, 3 Posts
2512.04797, cs․AI | cs․RO, 04 Dec 2025

🆕SIMA 2: A Generalist Embodied Agent for Virtual Worlds

SIMA team, Adrian Bolton, Alexander Lerchner, Alexandra Cordell, Alexandre Moufarek, Andrew Bolt, Andrew Lampinen, Anna Mitenkova, Arne Olav Hallingstad, Bo...
We introduce SIMA 2, a generalist embodied agent that understands and acts in a wide variety of 3D virtual worlds. Built upon a Gemini foundation model, SIMA 2 represents a significant step toward active, goal-directed interaction within an embodied environment. Unlike prior work (e.g., SIMA 1) limited to simple language commands, SIMA 2 acts as an interactive partner, capable of reasoning about high-level goals, conversing with the user, and handling complex instructions given through language and images. Across a diverse portfolio of games, SIMA 2 substantially closes the gap with human performance and demonstrates robust generalization to previously unseen environments, all while retaining the base model's core reasoning capabilities. Furthermore, we demonstrate a capacity for open-ended self-improvement: by leveraging Gemini to generate tasks and provide rewards, SIMA 2 can autonomously learn new skills from scratch in a new environment. This work validates a path toward creating versatile and continuously learning agents for both virtual and, eventually, physical worlds.
December 12, 2025 at 12:06 AM
paper.bsky.social
[25/30] 151 Likes, 43 Comments, 2 Posts
2512.08309, cs․CV | cs․AI | cs․GR | cs․LG, 09 Dec 2025

🆕Terrain Diffusion: A Diffusion-Based Successor to Perlin Noise in Infinite, Real-Time Terrain Generation

Alexander Goslin
For decades, procedural worlds have been built on procedural noise functions such as Perlin noise, which are fast and infinite, yet fundamentally limited in realism and large-scale coherence. We introduce Terrain Diffusion, an AI-era successor to Perlin noise that bridges the fidelity of diffusion models with the properties that made procedural noise indispensable: seamless infinite extent, seed-consistency, and constant-time random access. At its core is InfiniteDiffusion, a novel algorithm for infinite generation, enabling seamless, real-time synthesis of boundless landscapes. A hierarchical stack of diffusion models couples planetary context with local detail, while a compact Laplacian encoding stabilizes outputs across Earth-scale dynamic ranges. An open-source infinite-tensor framework supports constant-memory manipulation of unbounded tensors, and few-step consistency distillation enables efficient generation. Together, these components establish diffusion models as a practical foundation for procedural world generation, capable of synthesizing entire planets coherently, controllably, and without limits.
December 12, 2025 at 12:06 AM