SEO optimization - mainly meta descriptions

Author: lukeinglis

_pages/about.md

@@ -3,6 +3,7 @@ layout: page
 title: About
 permalink: /about/
 image: '/images/about/group.jpg'
+description: Meet the AI Innovation Team - researchers and engineers from Red Hat, IBM, MIT, and UMass building open-source LLM tools and advancing generative AI.
 ---
 
 ## AI Innovation Team

_pages/publications.md

@@ -3,6 +3,7 @@ layout: page
 title: Publications
 permalink: /publications/
 image:
+description: Research publications from the AI Innovation Team on LLM optimization, continual learning, synthetic data generation, and inference-time scaling.
 ---
 
 ## Select Publications

_pages/squat.md

@@ -2,6 +2,7 @@
 layout: page
 title: ""
 permalink: /squat/
+description: SQuat - KV cache quantization for scaling inference time compute in large language models. Optimize memory usage during LLM inference.
 ---
 
 <style>

_pages/videos.md

@@ -2,6 +2,7 @@
 layout: page
 title: Videos
 permalink: /videos/
+description: Watch our AI research talks, tutorials, and discussions on LLMs, machine learning, and generative AI. Featuring Random Samples seminars and No Math AI podcast.
 ---
 
 <div class="section-nav">

_posts/2025-02-05-r1-like-reasoning.md

@@ -4,6 +4,7 @@ title: Lessons on Reproducing R1-like Reasoning in Small LLMs without using Deep
 date: 2025-02-05
 image:  'https://probabilistic-inference-scaling.github.io/assets/images/process_video_cover.jpg'
 tags:   [reasoning, inference-time-scaling]
+description: Learn how to reproduce R1-like reasoning in small LLMs using particle filtering, synthetic data, and GRPO - achieving GPT-4o accuracy with only 4 rollouts.
 ---
 
 Written by Akash Srivastava, Isha Puri, Kai Xu, Shivchander Sudalairaj, Mustafa Eyceoz, Oleg Silkin, Abhishek Bhandwaldar, Aldo Genaro Pareja Cardona, GX Xu

_posts/2025-02-06-r1-like-reasoning-update-1.md

@@ -4,6 +4,7 @@ title: Update 1 - Lessons on Reproducing R1-like Reasoning in Small LLMs without
 date: 2025-02-06
 image:  'https://probabilistic-inference-scaling.github.io/assets/images/process_video_cover.jpg'
 tags:   [reasoning, inference-time-scaling]
+description: First update on R1-like reasoning experiments - Granite models show significant gains with particle filtering and new data quality experiments.
 ---
 
 Written by Akash Srivastava, Isha Puri, Kai Xu, Shivchander Sudalairaj, Mustafa Eyceoz, Oleg Silkin, Abhishek Bhandwaldar, Aldo Genaro Pareja Cardona, GX Xu

_posts/2025-02-07-r1-like-reasoning-update-2.md

@@ -4,6 +4,7 @@ title: Update 2 - Lessons on Reproducing R1-like Reasoning in Small LLMs without
 date: 2025-02-07
 image:  'https://probabilistic-inference-scaling.github.io/assets/images/process_video_cover.jpg'
 tags:   [reasoning, inference-time-scaling]
+description: Second update on R1 reasoning research - new results on training small LLMs with synthetic reasoning data and particle filtering methods.
 ---
 
 Written by Akash Srivastava, Isha Puri, Kai Xu, Shivchander Sudalairaj, Mustafa Eyceoz, Oleg Silkin, Abhishek Bhandwaldar, Aldo Genaro Pareja Cardona, GX Xu

_posts/2025-02-17-r1-like-reasoning-update-3.md

@@ -4,6 +4,7 @@ title: Update 3 - On Reasoning vs Inference-time scaling - Lessons on Reproducin
 date: 2025-02-17
 image:  'https://probabilistic-inference-scaling.github.io/assets/images/process_video_cover.jpg'
 tags:   [reasoning, inference-time-scaling]
+description: Understanding the distinction between reasoning and inference-time scaling in LLMs - insights from our R1 reproduction experiments.
 ---
 
 Written by Akash Srivastava, Isha Puri, Kai Xu, Shivchander Sudalairaj, Mustafa Eyceoz, Oleg Silkin, Abhishek Bhandwaldar, Aldo Genaro Pareja Cardona, GX Xu of the Red Hat AI Innovation Team

_posts/2025-04-04-orthogonal-subspace-learning.md

@@ -3,6 +3,7 @@ layout: post
 title: "Sculpting Subspaces: How We Solved Continual Learning in Large Language Models"
 date: 2025-04-04
 image:  '/images/posts/2025-04-04-orthogonal-subspace-learning/method_overview.png'
+description: Learn how our adaptive SVD method enables continual learning in LLMs with near-zero catastrophic forgetting, achieving 7% higher accuracy than baselines.
 ---
 
 **Authors:** Nikhil Shivakumar Nayak, Krishnateja Killamsetty, Ligong Han, Abhishek Bhandwaldar, Prateek Chanda, Kai Xu, Hao Wang, Aldo Pareja, Oleg Silkin, Mustafa Eyceoz, Akash Srivastava

_posts/2025-04-05-Async_GRPO_blog.md

@@ -3,6 +3,7 @@ layout: post
 title: Async-GRPO - Open, Fast, and Performant
 date: 2025-04-05
 image:  'https://hackmd.io/_uploads/HJrSol0ake.svg'
+description: Introducing Async-GRPO - an open-source library for scalable reinforcement learning with 42% efficiency gains over VERL and 11x over TRL for GRPO training.
 ---
 
 Written by Aldo Pareja, Mustafa Eyceoz