https://noisefree.news/ AI news with the noise removed — act/aware/watch signal, deterministic scrubbing, hourly refresh. en Thu, 05 Nov 2026 15:24:15 GMT Thu, 05 Nov 2026 15:24:15 GMT 60 https://www.reddit.com/r/MachineLearning/comments/1xol3vf/industry_to_academia_d https://www.reddit.com/r/MachineLearning/comments/1xol3vf/industry_to_academia_d Thu, 05 Nov 2026 15:11:22 GMT I have always had the desire to move to academia but I am not exactly sure how the process or profile building for academia works in general. Yeah, surely, I can ask AI and get answers but I don’t have much confidence in that . (Source excerpts) watch industry https://github.com/ggml-org/llama.cpp/releases/tag/b12128 https://github.com/ggml-org/llama.cpp/releases/tag/b12128 Thu, 05 Nov 2026 14:03:42 GMT tests: fix CI flakes of TOPK_MOE caused by near ties (#31426) * tests: fix CI flakes of TOPK_MOE caused by near ties With uniform random logits, two experts sometimes land within one f32 ulp of each other at a routing boundary, mostly with sqrt_softplus gating whose output range is narrow. (Source excerpts) Also covered by: GitHub Releases — https://github.com/ggml-org/llama.cpp/releases/tag/b12125 Also covered by: GitHub Releases — https://github.com/ggml-org/llama.cpp/releases/tag/b12124 aware open-source https://blog.google/company-news/outreach-and-initiatives/grow-with-google/national-veterans-small-business-week-2026 https://blog.google/company-news/outreach-and-initiatives/grow-with-google/national-veterans-small-business-week-2026 Thu, 05 Nov 2026 14:00:00 GMT Join our November 2026 YouTube livestream for National Veterans Small Business Week. Master AI skills with Gemini and get 1:1 coaching. (Source excerpts) aware industry https://www.reddit.com/r/MachineLearning/comments/1xohxyv/best_approach_for_demand_forecasting_across_40 https://www.reddit.com/r/MachineLearning/comments/1xohxyv/best_approach_for_demand_forecasting_across_40 Thu, 05 Nov 2026 12:55:51 GMT We're working with a synthetic retail supply chain dataset containing approximately 1.1 million records across 40 different locations . The goal is to predict product-level demand for each location as accurately as possible, so the forecasts can later be used by other agents for inventory planning and supply chain decisions. (Source excerpts) watch open-source https://www.reddit.com/r/MachineLearning/comments/1xofjrn/dialfork_a_benchmark_for_procedural_hallucination https://www.reddit.com/r/MachineLearning/comments/1xofjrn/dialfork_a_benchmark_for_procedural_hallucination Thu, 05 Nov 2026 10:44:15 GMT We recently released DialFork and thought it might interest people here. It's a benchmark for a failure mode of LLM agents that is easy to miss: every turn is fluent and factual, the tool calls are correct, yet the agent stops following the procedure it was given. (Source claim · not independently verified) aware model-release https://www.reddit.com/r/MachineLearning/comments/1xoc358/publication_expectations_for_ml_phds_at_top10_us https://www.reddit.com/r/MachineLearning/comments/1xoc358/publication_expectations_for_ml_phds_at_top10_us Thu, 05 Nov 2026 06:56:48 GMT For people at top-10 US CS schools or similarly competitive ML groups, is this roughly accurate? (Source excerpts) watch research https://latent.space/p/ainews-it-s-swapping-season https://latent.space/p/ainews-it-s-swapping-season Thu, 05 Nov 2026 06:24:09 GMT November 4, 2026 — OpenAI’s Linux kernel security work, Kimi K3.1 Pro’s open weights, and Ricursive’s Ripple-65B. (Source excerpts) aware open-source https://openai.com/index/setting-new-standard-responsible-ai-infrastructure https://openai.com/index/setting-new-standard-responsible-ai-infrastructure Thu, 05 Nov 2026 02:00:00 GMT OpenAI introduces its National Community Compact and Compute Campus Policy Blueprint to protect local resources, create jobs, and guide responsible AI infrastructure development. (Source excerpts) aware model-release https://github.com/openai/openai-python/releases/tag/v3.46.0 https://github.com/openai/openai-python/releases/tag/v3.46.0 Thu, 05 Nov 2026 00:55:06 GMT openai-python v3.46.0 (Headline only · read the source for details) Also covered by: GitHub Releases — https://github.com/openai/openai-python/releases/tag/v3.45.0 Also covered by: GitHub Releases — https://github.com/openai/openai-python/releases/tag/v3.44.0 watch open-source https://www.404media.co/american-voters-overwhelmingly-reject-ai-data-centers https://www.404media.co/american-voters-overwhelmingly-reject-ai-data-centers Thu, 05 Nov 2026 00:43:13 GMT American Voters Overwhelmingly Reject AI Data Centers (Headline only · read the source for details) Also covered by: Hacker News — https://news.ycombinator.com/item?id=50373505 watch industry https://github.com/langchain-ai/langchain/releases/tag/langchain-core%3D%3D1.7.6 https://github.com/langchain-ai/langchain/releases/tag/langchain-core%3D%3D1.7.6 Thu, 05 Nov 2026 00:22:12 GMT langchain langchain-core==1.7.6 (Headline only · read the source for details) Also covered by: GitHub Releases — https://github.com/langchain-ai/langchain/releases/tag/langchain-fireworks%3D%3D1.7.2 Also covered by: GitHub Releases — https://github.com/langchain-ai/langchain/releases/tag/langchain%3D%3D1.4.9 watch open-source https://orchidfiles.com/youll-need-nerves-of-steel-to-work-with-ai-agents https://orchidfiles.com/youll-need-nerves-of-steel-to-work-with-ai-agents Thu, 05 Nov 2026 00:01:30 GMT You'll need nerves of steel to work with AI agents (Headline only · read the source for details) Also covered by: Hacker News — https://news.ycombinator.com/item?id=50373149 watch industry https://huggingface.co/blog/phonon-tts https://huggingface.co/blog/phonon-tts Thu, 05 Nov 2026 00:00:00 GMT We announced Phonon in April and spent the next six months building on it with partners and early users. Their feedback helped us to shape this release. (Source excerpts) aware research https://www.federalregister.gov/documents/2026/11/05/2026-22593/medical-devices-hematology-and-pathology-devices-classification-of-the-digital-cervical-cytology https://www.federalregister.gov/documents/2026/11/05/2026-22593/medical-devices-hematology-and-pathology-devices-classification-of-the-digital-cervical-cytology Thu, 05 Nov 2026 00:00:00 GMT The special controls that apply to the device type are identified in this order and will be part of the codified language for classification of the digital cervical cytology slide imaging system with artificial intelligence algorithm. We are taking this action because we have determined that classifying the device into class II will provide a reasonable assurance of the safety and effectiveness of the device. (Source excerpts) aware policy https://www.federalregister.gov/documents/2026/11/05/2026-22615/proposed-priorities-and-definitions-secretarys-supplemental-priorities-and-definitions https://www.federalregister.gov/documents/2026/11/05/2026-22615/proposed-priorities-and-definitions-secretarys-supplemental-priorities-and-definitions Thu, 05 Nov 2026 00:00:00 GMT Department of Education (ED, or the Department) has priorities that represent the full spectrum of learners, reflect the expertise of the Federal agencies that are active partners in our mission, and incorporate lessons learned from implementation in FY 2026 grant competitions. (Source excerpts) aware policy https://www.aikido.dev/blog/cybersecurity-surviving-vulnerabilities-malware-best-practices https://www.aikido.dev/blog/cybersecurity-surviving-vulnerabilities-malware-best-practices Wed, 04 Nov 2026 21:06:00 GMT Login Menu EN Login Start for Free\\ \\ No CC required What your team can do to stay secure Written by Chris Holman Published on: Nov 4, 2026 It's been a brutal few weeks. A developer laptop led to the largest theft in crypto history at Binance. (Source excerpts) watch industry https://www.davidralphlewis.co.uk/posts/everything-is-an-ai-scam-now https://www.davidralphlewis.co.uk/posts/everything-is-an-ai-scam-now Wed, 04 Nov 2026 21:05:19 GMT Everything is an AI scam now (Headline only · read the source for details) Also covered by: Hacker News — https://news.ycombinator.com/item?id=50371508 watch incident https://github.com/anthropics/anthropic-sdk-python/releases/tag/v2.5.0 https://github.com/anthropics/anthropic-sdk-python/releases/tag/v2.5.0 Wed, 04 Nov 2026 20:27:20 GMT anthropic-sdk-python v2.5.0 (Headline only · read the source for details) watch open-source https://www.qualcomm.com/data-center/products/cloud-ai-100-ultra https://www.qualcomm.com/data-center/products/cloud-ai-100-ultra Wed, 04 Nov 2026 19:55:14 GMT Qualcomm Cloud AI 100 Ultra – 150W AI Accelerator Card with 128GB at 548GB/s (Headline only · read the source for details) Also covered by: Hacker News — https://news.ycombinator.com/item?id=50370654 watch industry https://latent.space/p/ainews-behind-latent-space-kill-my-saas-jev-ai-in-2026-aiei-shanghai https://latent.space/p/ainews-behind-latent-space-kill-my-saas-jev-ai-in-2026-aiei-shanghai Wed, 04 Nov 2026 19:17:30 GMT November 3, 2026 — A quiet day lets us share some behind the senes. (Source excerpts) aware industry https://arxiv.org/abs/2611.06914 https://arxiv.org/abs/2611.06914 Wed, 04 Nov 2026 18:58:23 GMT Researchers use large language models as agents in social simulations, where agents are prompted with observations and background information intended to shape their actions. Yet LLM agents have been shown to substantially change their behavior under meaning-preserving prompt variations, creating a validation problem for social simulation. (Source excerpts) aware research https://arxiv.org/abs/2611.06912 https://arxiv.org/abs/2611.06912 Wed, 04 Nov 2026 18:57:47 GMT FlowReconBench evaluates ten learned and five classical methods on three regimes: a laminar cylinder wake across Reynolds numbers, two-dimensional Kolmogorov turbulence and a large-scale three-dimensional (3D) isotropic turbulence. Methods share leakage-controlled splits and sensor draws and are scored on per-channel error, ensemble spread, calibration, spectra and cost. (Source excerpts) aware research https://arxiv.org/abs/2611.06911 https://arxiv.org/abs/2611.06911 Wed, 04 Nov 2026 18:57:17 GMT To fill this gap, we introduce BEAKER, a Benchmark for evaluating Embodied Actionability, Knowledge, Experimentation, and Reasoning in SDCLs. (Source claim · not independently verified) aware research https://arxiv.org/abs/2611.06910 https://arxiv.org/abs/2611.06910 Wed, 04 Nov 2026 18:56:22 GMT We introduce FrontierPhysics, an end-to-end benchmark with 56 tasks made by 42 physicists based on their own research, covering nine domains across theoretical, computational, and experimental physics. Each task requires agents to review the literature, develop a research plan, conduct research and experiments, and produce structured outputs and scientific manuscripts. (Source claim · not independently verified) aware research https://arxiv.org/abs/2611.06908 https://arxiv.org/abs/2611.06908 Wed, 04 Nov 2026 18:55:33 GMT We present HALO, a humanoid autonomy system powered by a pipeline that synthesizes vision-language-motion data across a thousand procedurally generated homes. The corpus scales with compute rather than human labor. (Source excerpts) aware research https://arxiv.org/abs/2611.06905 https://arxiv.org/abs/2611.06905 Wed, 04 Nov 2026 18:52:28 GMT The introduction of generalist robot policies has enabled a single, large model to learn to perform a variety of tasks from a vast pre-training corpus. While these models often exhibit powerful zero-shot generalization capabilities, they typically require an adaptation phase to achieve sufficient performance in deployment. (Source claim · not independently verified) aware research https://arxiv.org/abs/2611.06899 https://arxiv.org/abs/2611.06899 Wed, 04 Nov 2026 18:49:13 GMT We show that occlusion can be read off two distance estimates a camera-only pipeline already computes. Apparent height is largely unchanged by occlusion, because the detected box still spans the whole person; monocular metric depth is pulled toward the occluder, because the nearest surface inside the box is what a depth map reports. (Source excerpts) aware research https://arxiv.org/abs/2611.06888 https://arxiv.org/abs/2611.06888 Wed, 04 Nov 2026 18:41:40 GMT I examine how these conditions relate to occupational AI exposure using 21 O*NET items across 754 US occupations. Higher exposure is associated with lower condition scores under a large language model (LLM) measure and AI Occupational Exposure (AIOE), with equal occupation weights and employment weights. (Source excerpts) aware research https://arxiv.org/abs/2611.06879 https://arxiv.org/abs/2611.06879 Wed, 04 Nov 2026 18:35:35 GMT AMM (Adaptive Multi Mode) routes a known budget among quasi-Newton, trust-region, coordinate, multistart and evolution-strategy optimization engines with online probes of curvature, plateaus, separability, kinks, basins and fractal structure. Each probe has a default, a dimension-by-budget validity range and a cost charged to the budget. (Source excerpts) aware research https://research.google/blog/agile-deliberation-how-ai-helps-humans-define-the-gray-areas https://research.google/blog/agile-deliberation-how-ai-helps-humans-define-the-gray-areas Wed, 04 Nov 2026 18:07:05 GMT Agile Deliberation: How AI helps humans define the gray areas (Headline only · read the source for details) aware research https://huggingface.co/blog/Dharma-AI/dharma-field-extractor https://huggingface.co/blog/Dharma-AI/dharma-field-extractor Wed, 04 Nov 2026 17:50:10 GMT The model reads the page and works out which parts of it are structured fields (a label followed by a value, a table with a header row over its columns) from context and layout, not from a fixed list of field names, and returns everything it finds as key-value pairs. (Source excerpts) aware industry https://aws.amazon.com/blogs/machine-learning/four-layers-of-context-engineering-for-production-ai-agents-how-deep-insight-applies-them-in-working-code https://aws.amazon.com/blogs/machine-learning/four-layers-of-context-engineering-for-production-ai-agents-how-deep-insight-applies-them-in-working-code Wed, 04 Nov 2026 16:30:40 GMT This post presents a four-layer context engineering framework (architecture, prompts, file-based tools, and verification agents), implemented in working code from Deep Insight, an open source multi-agent data analysis solution on the Strands Agents SDK and Amazon Bedrock AgentCore. (Source excerpts) aware open-source https://aws.amazon.com/blogs/machine-learning/how-indegene-is-modernizing-and-scaling-enterprise-data-quality-in-life-sciences https://aws.amazon.com/blogs/machine-learning/how-indegene-is-modernizing-and-scaling-enterprise-data-quality-in-life-sciences Wed, 04 Nov 2026 16:25:57 GMT The agent generates context-aware data quality rules, validates them, and produces ready-to-run SQL and Python under human-in-the-loop governance, helping life sciences organizations scale enterprise data quality. (Source excerpts) aware industry https://aws.amazon.com/blogs/machine-learning/how-u-mobile-built-agentic-telco-operations-with-amazon-bedrock-agentcore https://aws.amazon.com/blogs/machine-learning/how-u-mobile-built-agentic-telco-operations-with-amazon-bedrock-agentcore Wed, 04 Nov 2026 16:23:33 GMT U Mobile, Malaysia's newest 5G network provider, built AiDaH, an agentic AI operations platform on Amazon Bedrock AgentCore that moves beyond AI assistance to autonomous action across telco operations. Nine months after production launch, AiDaH handles over 150 operational scenarios across six divisions, cutting trouble tickets by up to 84.5%. (Source claim · not independently verified) aware industry https://arxiv.org/abs/2611.06682 https://arxiv.org/abs/2611.06682 Wed, 04 Nov 2026 15:51:30 GMT We present \ours{}, a dynamic ontology-guided governance framework. \ours{} grounds heterogeneous calls in a shared semantic representation and combines an executable ontology with episode-level task state to derive deterministic ready, waiting, and forbidden decisions. (Source excerpts) aware research https://simonwillison.net/2026/Nov/4/datasette-llm https://simonwillison.net/2026/Nov/4/datasette-llm Wed, 04 Nov 2026 15:26:02 GMT Release: datasette-llm 0.1a10 Fixes a bug that affected my agent.datasette.io demo site, where the set of available LLM models was not correctly restricted to just those in the configuration. (Source excerpts) aware model-release https://arxiv.org/abs/2611.06629 https://arxiv.org/abs/2611.06629 Wed, 04 Nov 2026 15:12:30 GMT Using the universe of publicly disclosed vulnerabilities, I find that firms with access more than doubled their disclosures relative to Chinese and Russian tech firms without access, disclosed about 150 percent more high-impact vulnerabilities, and fixed reported flaws as quickly as before. (Source excerpts) aware research https://openai.com/index/openai-to-acquire-pierre https://openai.com/index/openai-to-acquire-pierre Wed, 04 Nov 2026 14:30:00 GMT OpenAI will acquire Pierre Computer Company, expanding Codex’s tools for building, understanding, and reviewing code. (Source excerpts) aware industry https://blog.google/innovation-and-ai/technology/health/evaluating-health-ai-at-scale https://blog.google/innovation-and-ai/technology/health/evaluating-health-ai-at-scale Wed, 04 Nov 2026 14:00:00 GMT Many of us have turned to AI with a health question or received care from a doctor who’s using it. (Source excerpts) aware policy https://arxiv.org/abs/2611.06257 https://arxiv.org/abs/2611.06257 Wed, 04 Nov 2026 10:07:14 GMT Open-weight language models that fit on a single consumer GPU are becoming capable on a growing range of tasks, and they matter because they can run locally without sending the code to a hosted service. (Source excerpts) aware research https://www.reddit.com/r/MachineLearning/comments/1xnjhdd/hydrogen10_openweight_scientific_model_beats_gpt6 https://www.reddit.com/r/MachineLearning/comments/1xnjhdd/hydrogen10_openweight_scientific_model_beats_gpt6 Wed, 04 Nov 2026 09:00:29 GMT We just released Hydrogen-1.0 , an open-weight model post-trained from GLM-5.3 for chemistry, materials science, and life sciences. Here’s how it performed in our evaluations: FrontierScience Research: 38.3% pass rate vs. (Source claim · not independently verified) watch model-release https://openai.com/index/a-thousand-fewer-paths-to-root https://openai.com/index/a-thousand-fewer-paths-to-root Wed, 04 Nov 2026 08:00:00 GMT OpenAI built an autonomous pipeline that found, validated, and generated patches for 948 local privilege-escalation vulnerabilities in the Linux kernel. (Source excerpts) aware incident https://openai.com/index/daybreak-red-aws-govcloud https://openai.com/index/daybreak-red-aws-govcloud Wed, 04 Nov 2026 07:00:00 GMT Daybreak Red brings GPT-5.6 Cyber to approved customers in AWS GovCloud (US-East) through Amazon Bedrock for authorized cybersecurity work. (Source excerpts) aware model-release https://simonwillison.net/2026/Nov/4/datasette-apps https://simonwillison.net/2026/Nov/4/datasette-apps Wed, 04 Nov 2026 00:29:36 GMT Release: datasette-apps 0.3a0 Datasette Apps adds support for hosting and editing custom HTML applications inside of Datasette. This release improves that integration: Apps can now be tagged in Datasette Agent chat messages by typing # , which autocompletes from the apps the user has permission to view. (Source claim · not independently verified) aware model-release https://openai.com/index/devrev-enterprise-self-service https://openai.com/index/devrev-enterprise-self-service Wed, 04 Nov 2026 00:00:00 GMT DevRev uses GPT-6.1 Sol to power Computer and Operator, balancing reasoning, speed, and cost for enterprise service. (Source excerpts) aware model-release https://openai.com/index/introducing-images-2-6 https://openai.com/index/introducing-images-2-6 Wed, 04 Nov 2026 00:00:00 GMT Images 2.6 brings faster image generation, GIF creation, more precise editing, and new API controls across ChatGPT, Codex, and the API. (Source claim · not independently verified) aware model-release https://github.com/ollama/ollama/releases/tag/v0.40.15-rc1 https://github.com/ollama/ollama/releases/tag/v0.40.15-rc1 Tue, 03 Nov 2026 22:52:31 GMT ollama v0.40.15-rc1 — v0.40.15 (Headline only · read the source for details) watch open-source https://simonwillison.net/2026/Nov/3/llm https://simonwillison.net/2026/Nov/3/llm Tue, 03 Nov 2026 21:41:42 GMT Release: llm 0.37.1 A bugfix release for LLM , with contributions from Steven Huang , Tao An , Arjan Mossel , and XQ . (Source excerpts) aware model-release https://huggingface.co/blog/allenai/statebench https://huggingface.co/blog/allenai/statebench Tue, 03 Nov 2026 17:15:21 GMT In code, this looks like let x = 4; in everyday life, it’s like a birth date—it stays fixed. Aperiodic problems allow reassignment by value. (Source excerpts) watch industry https://www.microsoft.com/en-us/research/podcast/learn-the-rules-the-hacker-mindset https://www.microsoft.com/en-us/research/podcast/learn-the-rules-the-hacker-mindset Tue, 03 Nov 2026 17:14:38 GMT The post Learn the rules: The hacker mindset appeared first on Microsoft Research . (Source excerpts) watch research https://deepmind.google/blog/agentic-cheminformatics-with-google-antigravity https://deepmind.google/blog/agentic-cheminformatics-with-google-antigravity Tue, 03 Nov 2026 17:01:49 GMT Learn how Google Antigravity and our Cheminformatics Skills accelerate drug discovery workflows, search compounds, and propose novel rigidified macrocycles. (Source excerpts) aware industry https://aws.amazon.com/blogs/machine-learning/servicenow-on-behalf-of-token-exchange-with-bedrock-agentcore-gateway https://aws.amazon.com/blogs/machine-learning/servicenow-on-behalf-of-token-exchange-with-bedrock-agentcore-gateway Tue, 03 Nov 2026 16:47:40 GMT Learn how to configure Amazon Bedrock AgentCore Gateway for an on-behalf-of (OBO) token exchange with Microsoft Entra ID, so an AI agent can call the ServiceNow REST API while preserving the calling user's identity. AgentCore Gateway performs the exchange natively, with no interceptor Lambda or custom code. (Source excerpts) watch industry https://aws.amazon.com/blogs/machine-learning/how-gen-digital-built-norton-revamps-ai-curated-news-engine-with-amazon-bedrock-agentcore-web-search https://aws.amazon.com/blogs/machine-learning/how-gen-digital-built-norton-revamps-ai-curated-news-engine-with-amazon-bedrock-agentcore-web-search Tue, 03 Nov 2026 16:43:20 GMT Gen Digital built Norton Revamp, a generative AI-powered professional reputation management platform, with an AI-curated News Engine grounded by Web Search on Amazon Bedrock AgentCore. See how the Gen team delivers real-time, cited content suggestions, goal setting, and news feeds with zero data egress and Norton-grade privacy. (Source excerpts) watch industry https://aws.amazon.com/blogs/machine-learning/build-a-semantic-layer-for-agentic-ai-with-edb-postgres-ai-and-amazon-bedrock-agentcore https://aws.amazon.com/blogs/machine-learning/build-a-semantic-layer-for-agentic-ai-with-edb-postgres-ai-and-amazon-bedrock-agentcore Tue, 03 Nov 2026 16:41:13 GMT Learn how to build a governed semantic layer for agentic AI inside EDB Postgres AI and query it from a Strands Agents agent on Amazon Bedrock AgentCore. Business meaning, join paths, and access control all live in Postgres, with GRANT and row-level security enforcing what each agent can read. (Source excerpts) watch industry https://aws.amazon.com/blogs/machine-learning/building-ultra-low-latency-voice-ai-applications-on-aws https://aws.amazon.com/blogs/machine-learning/building-ultra-low-latency-voice-ai-applications-on-aws Tue, 03 Nov 2026 16:40:54 GMT Learn how Cartesia, an AWS Partner, uses purpose-built State Space Models and AWS infrastructure to deliver sub-100ms text-to-speech and speech-to-text at global scale, and how to deploy its Sonic model on Amazon SageMaker AI or self-managed Amazon EC2 GPU instances. (Source excerpts) watch industry https://huggingface.co/blog/ibm-research/palette-deck-building https://huggingface.co/blog/ibm-research/palette-deck-building Tue, 03 Nov 2026 16:35:36 GMT Palette tests one bet: that a small open-weight model, trained on synthetic data in your design language and checked by a harness, can compete with frontier systems while you keep ownership of the result. We fine-tuned a 20B model to build decks the way people do: a designer plans the deck and a coder draws each slide. (Source excerpts) watch industry https://simonwillison.net/2026/Nov/3/qwen-can-interpret-model-weights https://simonwillison.net/2026/Nov/3/qwen-can-interpret-model-weights Tue, 03 Nov 2026 16:29:48 GMT The paper (OK, the authors classify is as a "blog post", but it's paper-shaped) found that frontier models, in particular GPT-6 Astra, can both interpret and modify the raw weights of a machine learning model - literally a text description of the model containing a bunch of floating point numbers. (Source excerpts) aware open-source