Jesse Thaler named director of the Laboratory for Nuclear Science
The professor of physics and inaugural director of the NSF AI Institute for Artificial Intelligence and Fundamental Interactions will lead LNS and continue his research in particle physics.
More from the blog
Measuring benchmark optimization in speech recognition
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
Multi-Vector (Late Interaction) Embedding Models with Sentence Transformers
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
Seeing beyond BMI: Estimating cardiometabolic risk with smartphone imagery
General Science
State of Open Models: Summer 2026 Observations
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
Related papers
Constitutional Midtraining: Content Presence Drives Alignment Gains
Post-training alignment is often shallow, eroding under fine-tuning. It remains untested as to whether constitutional midtraining interventions can produce durable alignment when cleanly isolated from post-training. We build a 394M-token constitutional corpus from Anthropic's Constitution and apply constitutional midtraining at 120B scale, where principled, values-based content is inserted into midtraining. A 2x2 design (curriculum ordering x deliberative reasoning) was used to produce four constitutionally midtrained conditions, plus a control, which were evaluated on self-generated and established benchmarks including alignment under pressure, value conflict resolution, blackmail, and emergent misalignment. All models were evaluated across three stages: post-midtraining, post-SFT, and post-benign fine-tuning. Constitutionally midtrained models outperformed the control on alignment generalization and durability, notably on blackmail: SFT instilled a blackmail propensity in all models, but constitutional midtraining blunted it, with the advantage surviving benign fine-tuning (-17.5pp). This durability did not extend to settings that required active resistance to in-context pressure or conflict, where the advantage attenuates after SFT. The presence of constitutional content at midtraining also mattered more than its structure, and constitutional midtraining incurred no capability cost, on average, at any stage (MMLU, ARC-Easy, piqa, GSM8K). A modest amount of constitutional content at midtraining could therefore yield broad, persistent alignment gains, offering a cheap, complementary addition to SFT-centered pipelines. Code, data, and models are available.
Professional Software Developers Don't Vibe, They Control: AI Agent Use for Coding in 2025
Investigating how experienced developers use agents in building software, including their motivations, strategies, task suitability, and sentiments finds that while experienced developers value agents as a productivity boost, they retain their agency in software design and implementation out of insistence on fundamental software quality attributes.
AI and Bullshit
It is argued that both AI and bullshitters are untrustworthy informants, and for similar reasons, it is natural to describe AI’s informational outputs as bullshit, as it signals their distinctive kind of epistemic deficiencies, which they share with bullshit.
Artificial Intelligence for Real-Time Cyber Threat Classification and Emerging Threat Detection: A Structured Review of Methods, Datasets, Challenges, and Research Directions
The reviewed literature indicates that AI-based methodologies often demonstrate superior detection capabilities for intricate and previously unseen attack patterns compared to traditional methods; however, direct performance comparisons are complicated due to discrepancies in datasets, experimental designs, and evaluation protocols.