Paper Feed: July 2026

Alignment and Control

• Diffuse AI Control on Fuzzy Tasks
  Mikhail Terekhov, Caglar Gulcehre, Vivek Hebbar, Joe Benton (2026)
• Sub-agent delegation chaining
  David Rein (2026)
• Data filtering works a lot worse than you would expect
  Dohun Lee, J Rosser, Josh Engels, Neel Nanda (2026)
• Fitness-seeking AIs
  Alex Mallen (2026)
• Distributed Attacks in Persistent-State AI Control
  Josh Hills, Ida Caspary, Asa Cooper Stickland (2026)
• Distributed Denial of Science: How Indirect Data Poisoning of AI Systems Can Industrialize Scientific Fraud
  Bálint Gyevnár, Atoosa Kasirzadeh, Nihar B. Shah (2026)
• Auditor-in-a-Box: Tools for Third-Party Auditing
  Roy Rinberg, Ben Penchas (2026)
• SOTA alignment assessments don't strongly update us against misalignment
  Alexa Pan (2026)
• Measuring Reward-Seeking via Contrastive Belief Updates
  Axel Højmark, Jérémy Scheurer, Evgenia Nitishinskaya, Felix Hofstätter, Jason Wolfe, Theodore Ehrenborg, Bronson Schoen, Alexander Meinke (2026)
• How our new Control Red Team is stress-testing frontier monitors
  AI Security Institute (2026)

AI Economics and Forecasting

• Toward an O*NET for AI R&D
  Jean-Stanislas Denain, Joe Kwon, Anson Ho (2026)
  See also: task taxonomy writeup
• Anthropic Economic Index report: Cadences
  Maxim Massenkoff, Eva Lyubich, Szymon Sacher, Zoe Hitzig, Shaoyi Zhang, Ryan Heller, Peter McCrory (2026)
• Expenditure Horizon: Measuring Optimization Ability, with an Application to NanoGPT
  Tom Cunningham, Manish Shetty, Vincent Cheng, Nate Rush (2026)
• The Economics of Recursive Self-Improvement
  Tom Cunningham, Lukas Althoff, Basil Halperin, Brian Jabarian, Andrew Koh, Arjun Ramani, Phil Trammell, Parker Whitfill, Cheryl Wu (2026)
• Contributions to OpenAI's Codex codebase show signs of AI uplift
  Jaeho Lee, Thomas Kwa (2026)
• LLM Detection as an Intervention: Downstream Impact under Strategic User Behavior
  Meena Jagadeesan, Tatsunori Hashimoto, Jon Kleinberg (2026)

Evaluations and Agents

• How bad data teaches models to write terrible code
  Max Niederman, Stephen Yang, Ege Erdil (2026)
• Smarter Spending: Dynamic Compute Allocation for Lean Theorem Provers
  Kari Rognvaldsson, Chenhao Sun, Jasper Dekoninck, Martin Vechev (2026)
• Should we benchmark conceptual capabilities using judgment prediction tasks?
  Alex Mallen (2026)
• CryptanalysisBench: Can LLMs do Cryptanalysis?
  Lukas Fluri, Avital Shafran, Nicholas Carlini, Matthew Jagielski, Milad Nasr, Orr Dunkelman, Eyal Ronen, Florian Tramèr (2026)
• Agents are under-elicited: A case study in optimization tasks
  Uzay Girit, Kaivalya Hariharan, Leni Shor, Rowan Huang (2026)
• Can AI agents conduct open-ended AI research? Early evidence from two case studies
  Peter Kirgis, Sayash Kapoor, Andrew Schwartz, Stephan Rabanser, David Africa, Konstantinos Voudouris, Viet Nguyen, Toby Pilditch, Magda Dubois, Harry Coppock, Cozmin Ududec, Nitya Nadgir, Matilda Orona, Tilman Bayer, Derrick Chan-Sew, Yue Ling, Abhishek Shetty, Helen Toner, Gillian Hadfield, Seth Lazar, Steve Newman, Shoshannah Tekofsky, Rishi Bommasani, Arvind Narayanan (2026)

Interpretability

• To what extent is Qwen3-32B predicting its persona?
  Arjun Khandelwal, Ryan Greenblatt, Alex Mallen (2026)
• LoRAcles: Self-Supervised Weight-Space Interpretability at Scale
  Celeste De Schamphelaere, Jan Philipp Bauer, Neel Nanda, Euan Ong (2026)
• Thousand-dimensional structure
  Geoffrey Irving, David Africa (2026)

Training and Generalization

• A (dis)analogy for (mis)understanding the lottery ticket hypothesis
  Vaishnavh Nagarajan (2026)
• Why Stochastic Gradient Descent Stops Just Short of the Edge
  Fangshuo Liao, Afroditi Kolomvaki, Anastasios Kyrillidis (2026)
  Paper: Stochastic Self-Stabilization
• Extractable Memorization From First Principles
  A. Feder Cooper, Marika Swanberg, Jamie Hayes, Lea Duesterwald, Christopher De Sa, Daniel E. Ho, Mark A. Lemley, Percy Liang (2026)
• Understanding Reasoning from Pretraining to Post-Training
  Jingyan Shen, Ang Li, Salman Rahman, Yifan Sun, Micah Goldblum, Matus Telgarsky, Pavel Izmailov (2026)
• Compression and Intelligence
  Ryan Greene (2026)

Miscellaneous

• What Language is This? Ask Your Tokenizer
  Clara Meister, Ahmetcan Yavuz, Pietro Lesci, Tiago Pimentel (2026)