bot: update RSS feed

feed.xml

@@ -7,7 +7,7 @@
     <docs>http://www.rssboard.org/rss-specification</docs>
     <generator>python-feedgen</generator>
     <language>en</language>
-    <lastBuildDate>Sun, 28 Dec 2025 00:02:46 +0000</lastBuildDate>
+    <lastBuildDate>Mon, 29 Dec 2025 00:02:47 +0000</lastBuildDate>
     <item>
       <title>GTR-Turbo: Merged Checkpoint is Secretly a Free Teacher for Agentic VLM Training</title>
       <link>https://arxiv.org/abs/2512.13043</link>
@@ -46,14 +46,14 @@
     <item>
       <title>Emergent temporal abstractions in autoregressive models enable hierarchical reinforcement learning</title>
       <link>https://arxiv.org/abs/2512.20605</link>
-      <description>&lt;p&gt;&lt;img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.20605.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /&gt;&lt;/p&gt;&lt;p&gt;&lt;b&gt;Authors:&lt;/b&gt; Seijin Kobayashi, Yanick Schimpf, Maximilian Schlegel, Angelika Steger, Maciej Wolczyk, Johannes von Oswald, Nino Scherrer, Kaitlin Maile, Guillaume Lajoie, Blake A. Richards, Rif A. Saurous, James Manyika, Blaise Agüera y Arcas, Alexander Meulemans, João Sacramento&lt;/p&gt;&lt;p&gt;&lt;b&gt;Upvotes:&lt;/b&gt; 43&lt;/p&gt;&lt;p&gt;&lt;b&gt;Summary:&lt;/b&gt; Large-scale autoregressive models pretrained on next-token prediction and finetuned with reinforcement learning (RL) have achieved unprecedented success on many problem domains. During RL, these models explore by generating new outputs, one token at a time. However, sampling actions token-by-token can result in highly inefficient learning, particularly when rewards are sparse. Here, we show that it is possible to overcome this problem by acting and exploring within the internal representations of an autoregressive model. Specifically, to discover temporally-abstract actions, we introduce a higher-order, non-causal sequence model whose outputs control the residual stream activations of a base autoregressive model. On grid world and MuJoCo-based tasks with hierarchical structure, we find that the higher-order model learns to compress long activation sequence chunks onto internal controllers. Critically, each controller executes a sequence of behaviorally meaningful actions that unfold over long timescales and are accompanied with a learned termination condition, such that composing multiple controllers over time leads to efficient exploration on novel tasks. We show that direct internal controller reinforcement, a process we term "internal RL", enables learning from sparse rewards in cases where standard RL finetuning fails. Our results demonstrate the benefits of latent action generation and reinforcement in autoregressive models, suggesting internal RL as a promising avenue for realizing hierarchical RL within foundation models.&lt;/p&gt;</description>
+      <description>&lt;p&gt;&lt;img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.20605.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /&gt;&lt;/p&gt;&lt;p&gt;&lt;b&gt;Authors:&lt;/b&gt; Seijin Kobayashi, Yanick Schimpf, Maximilian Schlegel, Angelika Steger, Maciej Wolczyk, Johannes von Oswald, Nino Scherrer, Kaitlin Maile, Guillaume Lajoie, Blake A. Richards, Rif A. Saurous, James Manyika, Blaise Agüera y Arcas, Alexander Meulemans, João Sacramento&lt;/p&gt;&lt;p&gt;&lt;b&gt;Upvotes:&lt;/b&gt; 47&lt;/p&gt;&lt;p&gt;&lt;b&gt;Summary:&lt;/b&gt; Large-scale autoregressive models pretrained on next-token prediction and finetuned with reinforcement learning (RL) have achieved unprecedented success on many problem domains. During RL, these models explore by generating new outputs, one token at a time. However, sampling actions token-by-token can result in highly inefficient learning, particularly when rewards are sparse. Here, we show that it is possible to overcome this problem by acting and exploring within the internal representations of an autoregressive model. Specifically, to discover temporally-abstract actions, we introduce a higher-order, non-causal sequence model whose outputs control the residual stream activations of a base autoregressive model. On grid world and MuJoCo-based tasks with hierarchical structure, we find that the higher-order model learns to compress long activation sequence chunks onto internal controllers. Critically, each controller executes a sequence of behaviorally meaningful actions that unfold over long timescales and are accompanied with a learned termination condition, such that composing multiple controllers over time leads to efficient exploration on novel tasks. We show that direct internal controller reinforcement, a process we term "internal RL", enables learning from sparse rewards in cases where standard RL finetuning fails. Our results demonstrate the benefits of latent action generation and reinforcement in autoregressive models, suggesting internal RL as a promising avenue for realizing hierarchical RL within foundation models.&lt;/p&gt;</description>
       <guid isPermaLink="false">https://arxiv.org/abs/2512.20605</guid>
       <pubDate>Tue, 23 Dec 2025 18:51:50 +0000</pubDate>
     </item>
     <item>
       <title>Latent Implicit Visual Reasoning</title>
       <link>https://arxiv.org/abs/2512.21218</link>
-      <description>&lt;p&gt;&lt;img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.21218.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /&gt;&lt;/p&gt;&lt;p&gt;&lt;b&gt;Authors:&lt;/b&gt; Kelvin Li, Chuyi Shang, Leonid Karlinsky, Rogerio Feris, Trevor Darrell, Roei Herzig&lt;/p&gt;&lt;p&gt;&lt;b&gt;Upvotes:&lt;/b&gt; 45&lt;/p&gt;&lt;p&gt;&lt;b&gt;Summary:&lt;/b&gt; While Large Multimodal Models (LMMs) have made significant progress, they remain largely text-centric, relying on language as their core reasoning modality. As a result, they are limited in their ability to handle reasoning tasks that are predominantly visual. Recent approaches have sought to address this by supervising intermediate visual steps with helper images, depth maps, or image crops. However, these strategies impose restrictive priors on what "useful" visual abstractions look like, add heavy annotation costs, and struggle to generalize across tasks. To address this critical limitation, we propose a task-agnostic mechanism that trains LMMs to discover and use visual reasoning tokens without explicit supervision. These tokens attend globally and re-encode the image in a task-adaptive way, enabling the model to extract relevant visual information without hand-crafted supervision. Our approach outperforms direct fine-tuning and achieves state-of-the-art results on a diverse range of vision-centric tasks -- including those where intermediate abstractions are hard to specify -- while also generalizing to multi-task instruction tuning.&lt;/p&gt;</description>
+      <description>&lt;p&gt;&lt;img src="https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2512.21218.png" alt="Paper thumbnail" style="max-width: 300px; height: auto;" /&gt;&lt;/p&gt;&lt;p&gt;&lt;b&gt;Authors:&lt;/b&gt; Kelvin Li, Chuyi Shang, Leonid Karlinsky, Rogerio Feris, Trevor Darrell, Roei Herzig&lt;/p&gt;&lt;p&gt;&lt;b&gt;Upvotes:&lt;/b&gt; 53&lt;/p&gt;&lt;p&gt;&lt;b&gt;Summary:&lt;/b&gt; While Large Multimodal Models (LMMs) have made significant progress, they remain largely text-centric, relying on language as their core reasoning modality. As a result, they are limited in their ability to handle reasoning tasks that are predominantly visual. Recent approaches have sought to address this by supervising intermediate visual steps with helper images, depth maps, or image crops. However, these strategies impose restrictive priors on what "useful" visual abstractions look like, add heavy annotation costs, and struggle to generalize across tasks. To address this critical limitation, we propose a task-agnostic mechanism that trains LMMs to discover and use visual reasoning tokens without explicit supervision. These tokens attend globally and re-encode the image in a task-adaptive way, enabling the model to extract relevant visual information without hand-crafted supervision. Our approach outperforms direct fine-tuning and achieves state-of-the-art results on a diverse range of vision-centric tasks -- including those where intermediate abstractions are hard to specify -- while also generalizing to multi-task instruction tuning.&lt;/p&gt;</description>
       <guid isPermaLink="false">https://arxiv.org/abs/2512.21218</guid>
       <pubDate>Wed, 24 Dec 2025 14:59:49 +0000</pubDate>
     </item>