Cybercab 공개 운행, 임베딩 압축, Diffusion 모델과 Coding Agent 구조

Tesla Robotaxi

@robotaxi

6시간 전

Cybercab, the people’s cab

Tesla AI

@Tesla_AI

6시간 전

FSD Supervised v14.3.9 starting to roll out shortly This release includes a new active safety feature set: FSD Supervised can now activate on your behalf when an imminent collision is detected and Automatic Emergency Braking (AEB) may not be enough. It may also engage if we detect you’re heavily distracted or have accidentally disengaged FSD

Peter Gostev

@petergostev

5시간 전

GPT-6-Astra (Max): Van Gogh Town in three.js. I want people to experience this, so I've now hosted it as well. It is pretty magical - turning 6 paintings by Van Gogh into a single walkable town, I love this. https:// van-goghs-town.surge.sh

5 embedding compression techniques, clearly explained: (bookmark this) Ten million 1,536-dimensional embeddings will occupy: - 62 GB in float32 - 15 GB in int8 - 2 GB as packed bits This only covers the raw vector payload, and an in-memory system also needs space for the ANN index, metadata, and allocator overhead. Embedding compression works along two axes: - the number of dimensions stored - the number of bits used for each dimension The five techniques in the visual reduce different parts of the payload. 1) PCA applies a post-training transform: It learns the directions with the most variance from a representative sample, then projects existing embeddings into a smaller space. It works with any embedding model, although the projection must be fitted and applied consistently to indexed vectors and queries. 2) MRL alters the training objective: The model is trained so that selected prefixes (say, the first <n> dimensions) of the embedding remain useful independently. You can therefore truncate an MRL embedding at inference without training a separate model for every target dimension. OpenAI reports that text-embedding-3-large at 256 dimensions still outperforms the 1,536-dimensional text-embedding-ada-002 on MTEB. 3-4) Scalar and binary quantization keep the dimension count fixed and reduce the representation used for each value. Scalar quantization typically maps float32 values to int8, giving a 4x reduction in the raw vector payload. The scale and offset add a small amount of metadata. Binary quantization keeps one bit per dimension, which gives a 32x reduction. Similarity search can then use XOR and a population count instead of floating-point distance calculations. 5) Product Quantization encodes subvectors as centroid IDs: It splits a vector into subvectors and replaces each subvector with the ID of its nearest centroid. Query distances are then approximated through centroid lookup tables. In all of these setups, the compressed representation does not need to produce the final ranking. Usually, you retrieve extra candidates from the compressed index, then recompute similarity using higher-precision document embeddings. This is especially useful with binary quantization. One bit can preserve enough coarse structure but loses magnitude information. Rescoring improves the ordering, but it cannot recover an item missed by the initial compressed-retrieval stage. Also, these methods can also work together. An MRL embedding can be truncated first and quantized afterward, reducing both dimension and precision. I built the binary retrieval pipeline end-to-end, including the rescoring pass. In the tested setup, it searches 36 million vectors in under 30 ms. Read it below.

LLaDA-Image: Building Strong Image Generators with Fully Open Training Recipes "We introduce LLaDA-Image, a unified framework that pairs a 6B Diffusion Transformer (DiT) trained from scratch with a frozen vision-language understanding module built on the LLaDA2.0-Mini diffusion language model backbone. Instead of relying heavily on paired image-text data from the beginning, we first build a strong visual generative prior through image-only pre-training and mid-training. The generation pipeline comprises 220M samples, 98% of which are real images. For efficient and scalable optimization, we use parameter-free RMSNorm throughout the DiT together with the Muon optimizer. The resulting unified model produces highly photorealistic images while accurately following fine-grained editing instructions. We further distill LLaDA-Image into LLaDA-Image-Turbo, enabling fast inference in 2-4 sampling steps. On Qwen-Image-Bench, LLaDA-Image achieves overall scores of 53.53 and 53.38 on the English and Chinese tracks, respectively, setting a new state-of-the-art among open-source models on both tracks. To support further research on capable and efficient generative models, we release our model weights, training code, and detailed recipes." abs: https:// arxiv.org/abs/2609.03796 code: https:// github.com/inclusionAI/LL aDA-Image … huggingface: https:// huggingface.co/inclusionAI/LL aDA-Image …

Unlocking Lossless Speedups in LLMs via Discrete Diffusion "we introduce diffusion-augmented LLMs, a new class of models that defines an AR model distribution while using diffusion to draw multiple tokens in parallel from that distribution." "Our 8B Uno model outperforms the leading open d-LLM, the 26B DiffusionGemma, and the proprietary Mercury 2 across all evaluated benchmarks in agentic tool use, coding, and long-context reasoning" website: https:// s-sahoo.com/uno/ arxiv: https:// arxiv.org/abs/2609.04010 code: https:// github.com/ifm-ai/uno huggingface: https:// huggingface.co/collections/s- sahoo/uno …

Hermes just made its own codebase dramatically cheaper for agents to understand The cleanup is not just for humans It directly reduces how much context agents burn while reading the repo. Some of the numbers are wild: • Median defining file size down 66% • Average lookup

Repo-To-Skill extracts the operational knowledge hidden inside codebases and converts it into compact, verified skills that other agents can reuse. GitHub may become more than a code source. https:// arxiv.org/html/2609.0274 9v1 …