Knowledge catalogue
Hey all! This started as a way to mark my 15th anniversary working in video game cinematics. I thought it would be fun to make a completely ridiculous, fictionalized version of how I got into the indu
DeepSeek Harness (dsh) is an open-source agent harness developed by DeepSeek AI. It uses an architecture where everything is a plugin, and is powered by Cordis, whose design is described in A Programm
🧩 DeepSeek Harness v0.1 is now available in Developer Preview! 🔹 We’re opening it up to developers building agent harnesses worldwide and open-sourcing the codebase in MIT license. 🔹 Powered by the Co
The Information: DeepSeek launches V4-Pro, its most advanced model that rivals Kimi K3 on some benchmarks but is priced much lower, at 0.435/1M input and 0.87/1M output tokens — Chinese AI developer D
arXiv:2608.11323v1 Announce Type: new Abstract: Enterprise practitioners read agent leaderboards as if they ranked agent capability. We show, across three open agent-trace benchmarks (TheAgentCompany,
arXiv:2608.11889v1 Announce Type: cross Abstract: Prompting-based (extit{i}.extit{e}., non-fine-tuning) Text-to-SQL methods, where underlying large language model parameters are not changed for the ta
arXiv:2608.12262v1 Announce Type: cross Abstract: Multimodal Large Language Models (MLLMs) have been growing the capability for scientific writing and collaboration. For example, OpenAI Prism is a fre
arXiv:2608.11249v1 Announce Type: cross Abstract: We study the problem of lossless text compression, motivated by the rapid growth in the collection and storage of digital textual data - including pla
arXiv:2608.11835v1 Announce Type: new Abstract: Semi-supervised video object segmentation is a challenging task that aims to segment a target throughout a video sequence given an initial mask at the f
arXiv:2608.11210v1 Announce Type: new Abstract: Bayesian calibration of process-based models requires a prior distribution for each model parameter. Despite decades of methodological work, researchers
arXiv:2608.11907v1 Announce Type: cross Abstract: As Large Vision-Language Models increasingly aim to integrate visual generation and understanding within a single parameter space, evaluating such str
arXiv:2608.12227v1 Announce Type: cross Abstract: Hyperspectral imaging (HSI) offers nondestructive assessment of fish freshness by detecting biochemical alterations across spectral bands. However, co
arXiv:2608.12308v1 Announce Type: cross Abstract: Aerial vision-language navigation (VLN) requires an embodied agent to integrate visual evidence over time, plan future actions, and determine when it
arXiv:2507.14267v2 Announce Type: replace Abstract: Large language model (LLM) agents can execute long-horizon scientific workflows, but their numerical outputs are difficult to trust: agents lose con
arXiv:2608.11690v1 Announce Type: new Abstract: Continual learning must absorb new tasks without erasing old ones, and replay---mixing a small buffer of past examples into current training---is among
Some people told me that the difference in richness and layout between Glimmer and Qwen wasn't clear to them. This example makes it super clear. I'm aware that comparing Glimmer 30B (a dense model) wi
arXiv:2608.11291v1 Announce Type: cross Abstract: Intrusion detection systems (IDS) and automated systems for detecting and reporting cyber threats, are commonly handled via supervised machine learnin
arXiv:2608.11584v1 Announce Type: new Abstract: Enterprise RAG deployments face a critical reliability gap: while LLMs satisfy 80% of individual constraints, only 26.8% of responses meet all requireme
arXiv:2608.11942v1 Announce Type: new Abstract: Quantitative MRI (qMRI) provides standardised tissue parameter maps, but the reliability of deep learning-based qMRI mapping methods is often not explic
arXiv:2509.16749v1 Announce Type: cross Abstract: LLMs are increasingly pervasive in the security environment, with limited measures of their effectiveness, which limits trust and usefulness to securi
Every document extraction system has a perception blind spot. We mapped them. For ExtractBench, we tested 14 systems on documents that weren't born digital: 1950s regulatory filings, hand-filled tax f
I managed to get Qwen3.8-2.4T-A95B running locally with llama.cpp on mu PC just for fun, cause why not. I was using the Unsloth Qwen3.8-2.4T-A95B-UD-Q1_0 GGUF quantization. The full GGUF is about 397
arXiv:2502.00837v3 Announce Type: replace-cross Abstract: Natural Language Processing (NLP) is now embedded in critical sectors including healthcare, finance, and customer relationship management, whe
arXiv:2608.11949v1 Announce Type: new Abstract: Roles provide an interpretable interface for organizing language-model agents, yet most multi-agent systems treat them as hand-written prompt labels dis
arXiv:2601.12178v2 Announce Type: replace Abstract: We propose a federated learning framework for the calibration of parametric insurance indices under heterogeneous renewable energy production losses
Qwen just released their first 3.8 model. The main addition in 3.8 is prompt-steered reasoning effort. You can tell the model how deeply to think by setting reasoning_effort to xhigh, medium, or low.
arXiv:2608.11868v1 Announce Type: new Abstract: The customization, optimization and stabilization of the process flow of a silicon bipolar phototransistor commits months of cleanroom time before a fin
arXiv:2608.11381v1 Announce Type: new Abstract: We study whether the localized numerical operations and integrative judgments of financial analysis benefit from the same form of LLM specialization. La
arXiv:2608.11562v1 Announce Type: cross Abstract: Videos captured through glass often contain reflections that degrade visual quality and interfere with downstream vision tasks. Although single-image
arXiv:2608.11683v1 Announce Type: new Abstract: AI agents are increasingly deployed for professional investment research, yet no benchmark captures the complexity of the full investor workflow. Existi
arXiv:2608.11739v1 Announce Type: cross Abstract: The prevailing recipe for Vision-Language-Action (VLA) models couples a pretrained VLM with a separately trained flow-matching action expert. This mak
arXiv:2608.11367v1 Announce Type: cross Abstract: Estimating human gaze targets from images in-the-wild is an important and formidable task. Existing approaches primarily employ brittle, multi-stage p
Gemini 3.7 Flash is newly released for the Devin platform. On FrontierCode 1.1 it matches Claude Sonnet 5‑level performance while costing less than half, and retains the series’ characteristic low lat
Ive been experimenting with task-aware GGUF quants for months, taking inspiration from TASA and TAQO but pushing the allocation lower to to the tensor level. The basic idea is to generate a custom ima
arXiv:2608.11263v1 Announce Type: new Abstract: Cross-modal place recognition (CMPR) aims to identify the same location across heterogeneous sensing modalities, such as vision and LiDAR. Existing meth
Overview ggml_cpu_fp16_to_fp32 leverages hardware F16C intrinsics (AVX-512, AVX2, etc.), faster than the software-only ggml_fp16_to_fp32_row, bringing 17-31% gain in prompt processing rate for a small
🧩 DeepSeek Harness v0.1 is now available in Developer Preview! 🔹 We’re opening it up to developers building agent harnesses worldwide and open-sourcing the codebase in MIT license. 🔹 Powered by the Co
arXiv:2608.11260v1 Announce Type: new Abstract: Video Anomaly Detection (VAD) aims to identify anomalous events and localize their temporal intervals. Existing approaches exhibit a 'when-what' dissoci
arXiv:2608.11332v1 Announce Type: new Abstract: Sign-language research for resource-constrained languages is often limited by the cost of dense linguistic labels such as glosses, temporal boundaries,
— Google AI Pro and Ultra subscribers can experience 3.7 Flash today via Spark in the @GeminiApp — Access the model in the Gemini Enterprise Agent Platform and Gemini Enterprise app — Build in the Gem
Google LLC today launched its most capable entry-level artificial intelligence model yet. Gemini 3.7 Flash is rolling out three weeks after its predecessor. Despite the short release cycle, Google eng
arXiv:2608.11344v1 Announce Type: cross Abstract: Financial institutions are delegating consequential decisions to agentic AI systems that decompose goals, coordinate models and tools, and act with li
Grok 4.6 just matched Claude Fable 5 on Perplexity’s WANDR benchmark....at more than 60% lower cost Both score exactly 0.496 But the cost per task is ridiculous: • Grok 4.6 → 7.58 • Claude Fable 5 → 2
Grok 4.6 ranks #1 on the GPQA Diamond leaderboard 🧠 Grok 4.6 (high) scores 95% - the highest score on the chart for graduate-level scientific reasoning It outperforms Claude Fable 5, Opus 5, GPT-5.6 S
arXiv:2608.11574v1 Announce Type: new Abstract: Hand Pose Estimation (HPE) is a fundamental technology for various applications such as AR/VR and robotics. In these applications, the visibility of eac
arXiv:2608.12122v1 Announce Type: cross Abstract: Robotic manipulation with dexterous hands is a cornerstone of Embodied AI, yet its progress is stifled by the high cost of collecting embodiment-aware
arXiv:2608.11224v1 Announce Type: new Abstract: Materials research advances through accumulated experience - scripts that work, protocols that are trusted, warnings attached to failed calculations or
arXiv:2608.11601v1 Announce Type: new Abstract: Driving world models are often interpreted as counterfactual simulators for observed driving episodes: given a factual driving log, they are asked what
arXiv:2608.11816v1 Announce Type: cross Abstract: State-aligned distortion has been documented in China-origin text-based large language models (LLMs), but whether, and in what form, it arises in mult
To my fellow crazies, the few. Those who dared wrestle with llama-70b, mistral-large, goliath, mistral8x22B, DeepSeekV2/3, wept when llama4 behemoth was announced, picked yourself up and are now wrest
arXiv:2608.12192v1 Announce Type: new Abstract: Foundation models for protein structure prediction remain unreliable on certain targets. External oracles can flag and correct these failures, but biolo
arXiv:2608.11692v1 Announce Type: new Abstract: Autonomous logistics sorting systems (ALSS) are an important industrial application of embodied AI, which requires joint planning over spatially disjoin
arXiv:2608.12194v1 Announce Type: cross Abstract: Kolmogorov-Arnold Networks (KANs) enhance nonlinear function approximation by replacing scalar weights with learnable univariate functions. However, a
Same old prompt, just appended a TIP in the end: 'Write a single HTML file with a full-page canvas and no libraries. Simulate a realistic side-view of a moving car as the main subject. Keep the car vi
arXiv:2608.12218v1 Announce Type: cross Abstract: Large language models are increasingly trained and deployed with long contexts that span documents, code repositories, and interaction histories. This
arXiv:2608.11234v1 Announce Type: new Abstract: Managing modern computing infrastructure has become a steadily harder problem due to the ever-increasing complexity. Recent advances in AI agents create
Hey everyone, Non-native speaker, writing my post by hand, let me know if I make mistakes (can only learn from it!) Muse Glimmer 30B is so far quite nice, but I haven't found a clear-cut case yet what
🦙☀️ It’s a hot Llama summer, and it pays to be a PRO at document parsing. For a limited time, upgrade to the LlamaParse Pro plan and we'll triple your credits: 400K → 1.2M credits in your first month.
arXiv:2608.11741v1 Announce Type: cross Abstract: The scholarly exegesis of ancient Chinese characters demands integrating visual observation, linguistic analysis, and historical context. However, exi
arXiv:2608.11786v1 Announce Type: new Abstract: Aggressive quantization disproportionately harms multilingual capability: in the sub-4B INT3 GPTQ regime, we measure 2-4x larger perplexity degradation