# Computer Science and Artificial Intelligence Laboratory (CSAIL)

The MIT Computer Science and Artificial Intelligence Laboratory (CSAIL) is an interdepartmental MIT laboratory that conducts research in computing and artificial intelligence.

This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles.

## MIT Researchers Used GPT-6 Astra To Edit Qwen's Open Weights Directly, Boosting Math Scores

DevFeed: [MIT Researchers Used GPT-6 Astra To Edit Qwen's Open Weights Directly, Boosting Math Scores](<https://devfeed.tech/articles/mit-researchers-used-gpt-6-astra-to-edit-qwen-s-open-weights-directly-boosting-math-scores-114833.md>)

Original publisher: [Read original article](<https://science.slashdot.org/story/26/11/03/0537252/mit-researchers-used-gpt-6-astra-to-edit-qwens-open-weights-directly-boosting-math-scores>)

Author: BeauHD

Published: 2026-11-03T16:00:00Z

Content type: news

Language: en

Sources: [Slashdot](<https://devfeed.tech/sources/slashdot.md>)

Topics: [Open Source Models & Datasets](<https://devfeed.tech/topics/open-source-models-datasets.md>), [gpt-6-astra](<https://devfeed.tech/topics/gpt-6-astra.md>), [qwen](<https://devfeed.tech/topics/qwen.md>), [Neural Network](<https://devfeed.tech/topics/neural-network.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Computer Science and Artificial Intelligence Laboratory (CSAIL)](<https://devfeed.tech/topics/computer-science-and-artificial-intelligence-laboratory-csail.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-systems](<https://devfeed.tech/tags/ai-systems.md>), [astra](<https://devfeed.tech/tags/astra.md>), [computer-science](<https://devfeed.tech/tags/computer-science.md>), [gpt](<https://devfeed.tech/tags/gpt.md>), [gpt-6](<https://devfeed.tech/tags/gpt-6.md>), [gpt-6-astra](<https://devfeed.tech/tags/gpt-6-astra.md>), [language-model](<https://devfeed.tech/tags/language-model.md>), [llm](<https://devfeed.tech/tags/llm.md>), [massachusetts-institute-of-technology](<https://devfeed.tech/tags/massachusetts-institute-of-technology.md>), [math](<https://devfeed.tech/tags/math.md>), [mit](<https://devfeed.tech/tags/mit.md>), [neural-networks](<https://devfeed.tech/tags/neural-networks.md>), [open](<https://devfeed.tech/tags/open.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [open-weights](<https://devfeed.tech/tags/open-weights.md>), [openai](<https://devfeed.tech/tags/openai.md>), [qwen](<https://devfeed.tech/tags/qwen.md>), [qwen3-5](<https://devfeed.tech/tags/qwen3-5.md>), [reasoning](<https://devfeed.tech/tags/reasoning.md>)

### AI overview

MIT CSAIL researchers report that GPT-6 Astra improved Qwen3.5-0.8B's MATH-500 score from 7.9% to 48.1% by directly editing about 0.016% of its weights, without conventional training procedures to generate the edits. They call this method "weight-space programming." The reported gain came primarily from correcting repetitive reasoning. The excerpt describes this as the strongest language-model example in broader experiments on using AI to inspect and alter neural networks; it does not provide further experimental details or establish how broadly the result generalizes.

### Source excerpt

An anonymous reader quotes a report from VentureBeat: Researchers from the Massachusetts Institute of Technology (MIT) Computer Science & Artificial Intelligence Laboratory (CSAIL) reported today that they used OpenAI's GPT-6 Astra language model to raise a small, open source Qwen model's math score roughly sixfold by directly editing its weights -- the numerical parameters that govern its behavior. Rikiya Takehi, Jayden Teoh, Alex Zhang and Omar Khattab described their work in a research blog post published November 2. They call the approach "weight-space programming" and tested Astra's ability to interpret and alter networks without using conventional training procedures to generate the edits. In the strongest language-model example, Astra's intervention raised Qwen3.5-0.8B's MATH-500 score from 7.9% to 48.1% while touching approximately 0.016% of its weights. The improvement came primarily from correcting repetitive reasoning--a distinction that shaped both the technical findings and the debate over their implications. Their experiments explored a broader possibility: AI systems that inspect, repair and construct other neural networks by working directly on their internal machinery. "This is cool as heck!" wrote Yuchen Jin, a technical staffer at Databricks and former co-founder and CTO of Hyperbolic Labs who was unaffiliated with the research, on X. "Imagine using a large LLM to directly edit a smaller model's weights and dramatically improve its capabilities. It's like gene editing for neural networks. Even cooler than distillation!" Voyage AI staff researcher Chandan Singh, formerly of Microsoft AI, also reacted to the findings: "For anyone questioning whether this post is overhyped, I think they are actually *underselling* the approach." Read more of this story at Slashdot.

## US government, tech giants and Biohub commit $1.8B to AI biology initiative

DevFeed: [US government, tech giants and Biohub commit $1.8B to AI biology initiative](<https://devfeed.tech/articles/us-government-tech-giants-and-biohub-commit-1-8b-to-ai-biology-initiative-67070.md>)

Original publisher: [Read original article](<https://siliconangle.com/2026/10/07/us-government-tech-giants-and-biohub-commit-1-8b-to-ai-biology-initiative/>)

Author: Maria Deutscher

Published: 2026-10-07T22:46:47Z

Content type: news

Language: en

Sources: [SiliconANGLE](<https://devfeed.tech/sources/siliconangle.md>)

Topics: [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [BioNeMo](<https://devfeed.tech/topics/bionemo.md>), [Computer Science and Artificial Intelligence Laboratory (CSAIL)](<https://devfeed.tech/topics/computer-science-and-artificial-intelligence-laboratory-csail.md>), [Frontier AI](<https://devfeed.tech/topics/frontier-ai.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [biohub](<https://devfeed.tech/tags/biohub.md>), [biology](<https://devfeed.tech/tags/biology.md>), [deepmind](<https://devfeed.tech/tags/deepmind.md>), [emerging-tech](<https://devfeed.tech/tags/emerging-tech.md>), [google-plus-deepmind](<https://devfeed.tech/tags/google-plus-deepmind.md>), [government](<https://devfeed.tech/tags/government.md>), [isomorphic-labs](<https://devfeed.tech/tags/isomorphic-labs.md>), [meta](<https://devfeed.tech/tags/meta.md>), [news](<https://devfeed.tech/tags/news.md>), [research](<https://devfeed.tech/tags/research.md>), [the-latest](<https://devfeed.tech/tags/the-latest.md>), [training-dataset](<https://devfeed.tech/tags/training-dataset.md>), [virtual-cell](<https://devfeed.tech/tags/virtual-cell.md>)

### AI overview

A public and private sector consortium is committing $1.8 billion to develop biology datasets for AI models. The U.S. Department of Energy, Biohub, NIH, Google DeepMind, Isomorphic Labs, Meta and Nvidia are among the participants, contributing funding, datasets, microscopy equipment, accelerators or research software. The initiative aims to support biology-optimized models and simulated cells that could help scientists study therapies and conduct experiments digitally.

### Source excerpt

More than a half-dozen public and private sector organizations are teaming up to develop biology datasets for artificial intelligence models. The consortium, which announced the collaboration today, is committing $1.8 billion to the project. Studying how a cell responds to a new therapy usually requires scientists to set up a physical test environment. That can [...] The post US government, tech giants and Biohub commit $1.8B to AI biology initiative appeared first on SiliconANGLE.

## When AI art has no author: Study finds generated images often can't be traced to training data

DevFeed: [When AI art has no author: Study finds generated images often can't be traced to training data](<https://devfeed.tech/articles/when-ai-art-has-no-author-study-finds-generated-images-often-can-t-be-traced-to-training-data-37986.md>)

Original publisher: [Read original article](<https://news.mit.edu/2026/when-ai-art-has-no-author-generated-images-often-cant-be-traced-to-training-data-0818>)

Author: Rachel Gordon | MIT CSAIL

Published: 2026-08-18T16:35:00Z

Content type: news

Language: en

Sources: [MIT AI News](<https://devfeed.tech/sources/mit-ai-news.md>)

Topics: [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [AI Research](<https://devfeed.tech/topics/ai-research.md>), [dataset](<https://devfeed.tech/topics/dataset.md>), [Computer Science and Artificial Intelligence Laboratory (CSAIL)](<https://devfeed.tech/topics/computer-science-and-artificial-intelligence-laboratory-csail.md>)

Tags: [ablation](<https://devfeed.tech/tags/ablation.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-and-copyright-law](<https://devfeed.tech/tags/ai-and-copyright-law.md>), [ai-generated-images](<https://devfeed.tech/tags/ai-generated-images.md>), [ai-governance](<https://devfeed.tech/tags/ai-governance.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [arts](<https://devfeed.tech/tags/arts.md>), [arts-technology-and-society](<https://devfeed.tech/tags/arts-technology-and-society.md>), [attribution-decay](<https://devfeed.tech/tags/attribution-decay.md>), [causal-inference](<https://devfeed.tech/tags/causal-inference.md>), [computer-science-and-artificial-intelligence-laboratory-csail](<https://devfeed.tech/tags/computer-science-and-artificial-intelligence-laboratory-csail.md>), [computer-science-and-technology](<https://devfeed.tech/tags/computer-science-and-technology.md>), [counterfactual-analysis](<https://devfeed.tech/tags/counterfactual-analysis.md>), [counterfactual-radius](<https://devfeed.tech/tags/counterfactual-radius.md>), [data](<https://devfeed.tech/tags/data.md>), [data-attribution](<https://devfeed.tech/tags/data-attribution.md>), [datasets](<https://devfeed.tech/tags/datasets.md>), [david-gifford](<https://devfeed.tech/tags/david-gifford.md>), [diffusion-ensembles](<https://devfeed.tech/tags/diffusion-ensembles.md>), [electrical-engineering-and-computer-science-eecs](<https://devfeed.tech/tags/electrical-engineering-and-computer-science-eecs.md>), [ethics](<https://devfeed.tech/tags/ethics.md>), [generative-ai-images](<https://devfeed.tech/tags/generative-ai-images.md>), [generative-diffusion-models](<https://devfeed.tech/tags/generative-diffusion-models.md>), [image-similarity-metrics](<https://devfeed.tech/tags/image-similarity-metrics.md>), [images](<https://devfeed.tech/tags/images.md>), [law](<https://devfeed.tech/tags/law.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [machine-unlearning](<https://devfeed.tech/tags/machine-unlearning.md>), [mit-csail](<https://devfeed.tech/tags/mit-csail.md>), [mit-schwarzman-college-of-computing](<https://devfeed.tech/tags/mit-schwarzman-college-of-computing.md>), [model](<https://devfeed.tech/tags/model.md>), [model-interpretability](<https://devfeed.tech/tags/model-interpretability.md>), [paper](<https://devfeed.tech/tags/paper.md>), [privacy-preserving-machine-learning](<https://devfeed.tech/tags/privacy-preserving-machine-learning.md>), [research](<https://devfeed.tech/tags/research.md>), [school-of-engineering](<https://devfeed.tech/tags/school-of-engineering.md>), [science](<https://devfeed.tech/tags/science.md>), [technology-and-policy](<https://devfeed.tech/tags/technology-and-policy.md>), [technology-and-society](<https://devfeed.tech/tags/technology-and-society.md>), [training-data-attribution](<https://devfeed.tech/tags/training-data-attribution.md>), [training-data-influence](<https://devfeed.tech/tags/training-data-influence.md>), [zheng-dai](<https://devfeed.tech/tags/zheng-dai.md>)

### AI overview

MIT CSAIL researchers describe attribution decay, a phenomenon in which the influence of individual training examples on a generative model's outputs diminishes as datasets grow. Their method removes training examples and retrains models to test whether generated samples change.

### Source excerpt

A new method for surgically removing training examples from a model reveals that as datasets grow, the link between what a model learns and what it produces dissolves.

## GeoPT helps AI models simulate how objects respond to physical forces

DevFeed: [GeoPT helps AI models simulate how objects respond to physical forces](<https://devfeed.tech/articles/with-a-feel-for-physics-ai-models-simulate-a-wider-range-of-real-world-scenarios-37942.md>)

Original publisher: [Read original article](<https://news.mit.edu/2026/ai-models-simulate-wider-range-of-real-world-scenarios-0810>)

Author: Alex Shipps | MIT CSAIL

Published: 2026-08-10T19:25:00Z

Content type: news

Language: en

Sources: [MIT AI News](<https://devfeed.tech/sources/mit-ai-news.md>)

Topics: [AI Models](<https://devfeed.tech/topics/ai-models.md>), [Simulation](<https://devfeed.tech/topics/simulation.md>), [AI Research](<https://devfeed.tech/topics/ai-research.md>), [Computer Science and Artificial Intelligence Laboratory (CSAIL)](<https://devfeed.tech/topics/computer-science-and-artificial-intelligence-laboratory-csail.md>), [Robotics](<https://devfeed.tech/topics/robotics.md>)

Tags: [3-d-imaging](<https://devfeed.tech/tags/3-d-imaging.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-models](<https://devfeed.tech/tags/ai-models.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [computational-fluid-dynamics-cfd](<https://devfeed.tech/tags/computational-fluid-dynamics-cfd.md>), [computer-graphics](<https://devfeed.tech/tags/computer-graphics.md>), [computer-modeling](<https://devfeed.tech/tags/computer-modeling.md>), [computer-science-and-artificial-intelligence-laboratory-csail](<https://devfeed.tech/tags/computer-science-and-artificial-intelligence-laboratory-csail.md>), [computer-science-and-technology](<https://devfeed.tech/tags/computer-science-and-technology.md>), [crash-simulation](<https://devfeed.tech/tags/crash-simulation.md>), [design](<https://devfeed.tech/tags/design.md>), [electrical-engineering-and-computer-science-eecs](<https://devfeed.tech/tags/electrical-engineering-and-computer-science-eecs.md>), [geometric-pre-training](<https://devfeed.tech/tags/geometric-pre-training.md>), [geopt](<https://devfeed.tech/tags/geopt.md>), [haixu-wu](<https://devfeed.tech/tags/haixu-wu.md>), [human-computer-interaction](<https://devfeed.tech/tags/human-computer-interaction.md>), [kaiming-he](<https://devfeed.tech/tags/kaiming-he.md>), [manufacturing](<https://devfeed.tech/tags/manufacturing.md>), [minghao-guo](<https://devfeed.tech/tags/minghao-guo.md>), [mit-csail](<https://devfeed.tech/tags/mit-csail.md>), [mit-eecs](<https://devfeed.tech/tags/mit-eecs.md>), [mit-schwarzman-college-of-computing](<https://devfeed.tech/tags/mit-schwarzman-college-of-computing.md>), [neural-physics-simulation](<https://devfeed.tech/tags/neural-physics-simulation.md>), [paper](<https://devfeed.tech/tags/paper.md>), [physics](<https://devfeed.tech/tags/physics.md>), [physics-aware-ai](<https://devfeed.tech/tags/physics-aware-ai.md>), [physics-foundation-models](<https://devfeed.tech/tags/physics-foundation-models.md>), [research](<https://devfeed.tech/tags/research.md>), [robotics](<https://devfeed.tech/tags/robotics.md>), [school-of-engineering](<https://devfeed.tech/tags/school-of-engineering.md>), [self-supervised-learning](<https://devfeed.tech/tags/self-supervised-learning.md>), [simulation](<https://devfeed.tech/tags/simulation.md>), [surrogate-modeling](<https://devfeed.tech/tags/surrogate-modeling.md>), [synthetic-dynamics](<https://devfeed.tech/tags/synthetic-dynamics.md>), [transformer-based-simulators](<https://devfeed.tech/tags/transformer-based-simulators.md>), [wojciech-matusik](<https://devfeed.tech/tags/wojciech-matusik.md>)

### AI overview

Researchers at MIT CSAIL and Tsinghua University developed GeoPT, a pre-training approach that uses 3D simulations of mechanical interactions to help AI models learn physics more efficiently. The article reports that models using the approach reached peak performance twice as fast and trained on up to 60 percent less data than leading models.

### Source excerpt

"GeoPT" helps AI models understand the basics of physics so they can simulate how objects respond to things like wind and water more efficiently and accurately.

## Tiny robot boats build floating structures

DevFeed: [Tiny robot boats build floating structures](<https://devfeed.tech/articles/tiny-robot-boats-build-floating-structures-37983.md>)

Original publisher: [Read original article](<https://news.mit.edu/2026/tiny-robot-boats-build-floating-structures-0709>)

Author: Rachel Gordon | MIT CSAIL

Published: 2026-07-09T15:50:00Z

Content type: news

Language: en

Sources: [MIT AI News](<https://devfeed.tech/sources/mit-ai-news.md>)

Topics: [Robotics](<https://devfeed.tech/topics/robotics.md>), [Computer Science and Artificial Intelligence Laboratory (CSAIL)](<https://devfeed.tech/topics/computer-science-and-artificial-intelligence-laboratory-csail.md>), [Computer science](<https://devfeed.tech/topics/computer-science.md>)

Tags: [alejandro-gonzalez-garcia](<https://devfeed.tech/tags/alejandro-gonzalez-garcia.md>), [aquatic-robots](<https://devfeed.tech/tags/aquatic-robots.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [autonomous-vehicles](<https://devfeed.tech/tags/autonomous-vehicles.md>), [bioinspiration](<https://devfeed.tech/tags/bioinspiration.md>), [cities](<https://devfeed.tech/tags/cities.md>), [computer-science-and-artificial-intelligence-laboratory-csail](<https://devfeed.tech/tags/computer-science-and-artificial-intelligence-laboratory-csail.md>), [computer-science-and-technology](<https://devfeed.tech/tags/computer-science-and-technology.md>), [daniela-rus](<https://devfeed.tech/tags/daniela-rus.md>), [electrical-engineering-and-computer-science-eecs](<https://devfeed.tech/tags/electrical-engineering-and-computer-science-eecs.md>), [fire-ants](<https://devfeed.tech/tags/fire-ants.md>), [floatform](<https://devfeed.tech/tags/floatform.md>), [floating-infrastructure](<https://devfeed.tech/tags/floating-infrastructure.md>), [hybrid-coordination](<https://devfeed.tech/tags/hybrid-coordination.md>), [infrastructure](<https://devfeed.tech/tags/infrastructure.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [mit-csail](<https://devfeed.tech/tags/mit-csail.md>), [mit-eecs](<https://devfeed.tech/tags/mit-eecs.md>), [mit-schwarzman-college-of-computing](<https://devfeed.tech/tags/mit-schwarzman-college-of-computing.md>), [mit-sea-grant](<https://devfeed.tech/tags/mit-sea-grant.md>), [mit-sea-grant-program](<https://devfeed.tech/tags/mit-sea-grant-program.md>), [mit-senseable-city-lab](<https://devfeed.tech/tags/mit-senseable-city-lab.md>), [modular-robotic-boats](<https://devfeed.tech/tags/modular-robotic-boats.md>), [modular-self-reconfigurable-robots](<https://devfeed.tech/tags/modular-self-reconfigurable-robots.md>), [niklas-hagemann](<https://devfeed.tech/tags/niklas-hagemann.md>), [nonlinear-hydrodynamics](<https://devfeed.tech/tags/nonlinear-hydrodynamics.md>), [raft](<https://devfeed.tech/tags/raft.md>), [research](<https://devfeed.tech/tags/research.md>), [robot-boats](<https://devfeed.tech/tags/robot-boats.md>), [robot-self-assembly](<https://devfeed.tech/tags/robot-self-assembly.md>), [robotics](<https://devfeed.tech/tags/robotics.md>), [robots](<https://devfeed.tech/tags/robots.md>), [school-of-architecture-and-planning](<https://devfeed.tech/tags/school-of-architecture-and-planning.md>), [school-of-engineering](<https://devfeed.tech/tags/school-of-engineering.md>), [science](<https://devfeed.tech/tags/science.md>), [self-reconfiguration](<https://devfeed.tech/tags/self-reconfiguration.md>), [senseable-city-lab](<https://devfeed.tech/tags/senseable-city-lab.md>), [snap](<https://devfeed.tech/tags/snap.md>), [swarm](<https://devfeed.tech/tags/swarm.md>), [swarm-robots](<https://devfeed.tech/tags/swarm-robots.md>), [tiny](<https://devfeed.tech/tags/tiny.md>), [transportation](<https://devfeed.tech/tags/transportation.md>), [ultrasonic-beacon-system](<https://devfeed.tech/tags/ultrasonic-beacon-system.md>), [urban-studies-and-planning](<https://devfeed.tech/tags/urban-studies-and-planning.md>), [water](<https://devfeed.tech/tags/water.md>), [wei-wang](<https://devfeed.tech/tags/wei-wang.md>)

### AI overview

MIT researchers developed FloatForm, a swarm of small robotic boats that can assemble into reconfigurable structures on the water and later break apart and reassemble into new configurations with minimal human direction.

### Source excerpt

MIT researchers developed FloatForm, a swarm of small aquatic robots that snap together like ants forming a raft, assembling into reconfigurable structures on the water.