HN user

badmonster

426 karma
Posts115
Comments233
View on HN
arxiv.org 7mo ago

Selfi: Self Improving Reconstruction Engine via 3D Geometric Feature Alignment

badmonster
2pts0
arxiv.org 7mo ago

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

badmonster
2pts0
arxiv.org 7mo ago

Agentic Learner with Grow-and-Refine Multimodal Semantic Memory

badmonster
1pts0
cocoindex.io 7mo ago

Extract structured information from Hacker News and keep in sync with Postgres

badmonster
2pts0
arxiv.org 7mo ago

When Video Detailed Captioners Evolve Themselves via Agentic Self-Reflection

badmonster
2pts1
cocoindex.io 8mo ago

Streamline Structured and Unstructured Data Flows from Postgres with AI

badmonster
1pts0
arxiv.org 8mo ago

Counterfactual World Models via Digital Twin-Conditioned Video Diffusion

badmonster
2pts0
cocoindex.io 8mo ago

How to perform adaptive batching for massive remote LLM calls

badmonster
2pts0
arxiv.org 8mo ago

Continuous Autoregressive Language Models

badmonster
2pts0
cocoindex.io 8mo ago

Adaptive Batching – 5x throughput on your data pipelines

badmonster
1pts0
research.google 8mo ago

Nested Learning – A new ML paradigm for continual learning

badmonster
7pts0
arxiv.org 9mo ago

Grasp Any Region: Precise, Contextual Pixel Understanding for Multimodal LLMs

badmonster
1pts0
arxiv.org 9mo ago

StreamingVLM: Real-Time Understanding for Infinite Video Streams

badmonster
33pts0
cocoindex.io 9mo ago

Automated invoice processing with AI and incremental processing

badmonster
1pts0
arxiv.org 10mo ago

Outcome-Based Exploration for LLM Reasoning

badmonster
2pts0
cocoindex.io 10mo ago

Incrementally Transform Structured and Unstructured Data from Postgres with AI

badmonster
1pts0
arxiv.org 11mo ago

A Systematic Study of Post-Training Quantization for Diffusion LLMs

badmonster
1pts0
cocoindex.io 11mo ago

Multi-Format RAG with ColPali: PDFs, Images, Slides – No OCR

badmonster
2pts0
arxiv.org 11mo ago

Mathematical Computation and Reasoning Errors by Large Language Models

badmonster
2pts0
cocoindex.io 11mo ago

Multi-Dimensional Vector Support in CocoIndex – Underneath Explained

badmonster
1pts1
research.google 11mo ago

Achieving 10,000x training data reduction with high-fidelity labels

badmonster
154pts29
arxiv.org 11mo ago

Rethinking Analytical Processing in the GPU Era

badmonster
2pts0
cocoindex.io 11mo ago

Thinking in Rust: Ownership, Access, and Memory Safety

badmonster
2pts0
cocoindex.io 11mo ago

Indexing Faces for Scalable Visual Search – Build Google Photo Search

badmonster
1pts0
techcrunch.com 12mo ago

Sam Altman warns there's no legal confidentiality using ChatGPT as a therapist

badmonster
11pts6
research.google 12mo ago

LSM-2: Learning from incomplete wearable sensor data

badmonster
4pts1
arxiv.org 12mo ago

Explainable Mapper: Charting LLM Embedding Spaces Using Perturbation-Based

badmonster
2pts1
arxiv.org 1y ago

Diffusion Beats Autoregressive in Data-Constrained Settings

badmonster
3pts1
cocoindex.io 1y ago

Realtime codebase indexing for coding agents with – 50 lines of Python

badmonster
2pts1
arxiv.org 1y ago

Impact of Pretraining Word Co-Occurrence on Compositional Generalization In

badmonster
2pts1

"Code red" feels like theater. Competition is healthy - Google's compute advantage was always going to matter once they got serious. The real question isn't who's ahead this quarter, but whether anyone can maintain a moat when the underlying tech is rapidly commoditizing.

Show HN: RFC Hub 8 months ago

What RFC standards or protocols can users browse on RFC Hub? Does it include IETF RFCs or is it focused on internal organizational RFCs?

Notes on Bhutan 8 months ago

Interesting parallel. Does this reliance limit Bhutan's sovereignty in practice? What's the trade-off?

Do these behavioral differences have performance implications? Which approach is more efficient in practice?

Hi I've been working on adding multi-vector support natively in cocoindex for multi-modal RAG at scale. I wrote blog to help understand the concept of multi-vector and how it works underneath.

The framework itself automatically infers types, so when defining a flow, we don’t need to explicitly specify any types. Felt these concept are fundamental to multimodal data processing so just wanted to share. This unlocks at scale: images, text, audio, video — all can be represented as structured multi-vectors that preserve the unique semantics of each modality.

breakdown + Python examples: https://cocoindex.io/blogs/multi-vector/ Star GitHub if you like it! https://github.com/cocoindex-io/cocoindex