16 November 2026
Boule or Baguette?
A Study on Task Topology, Length Generalization, and the Benefit of Reasoning Traces
By: Willliam L. Tong and Ege Çakar
Reasoning traces are central to how modern AI models solve hard problems, but it remains unclear when they actually help. On a new dataset of…
9 November 2026
Unlocking Lossless Speedups in LLMs via Discrete Diffusion
By: Jonathan Geuter and Subham Sahoo
Large language models generate text one token at a time, which makes inference slow. Diffusion language models generate many tokens in parallel, but they lose quality, and their speedups fade when a server handles many requests at once. We introduce Uno: a standard autoregressive (AR) model plus small diffusion weights that draft a block of tokens in one forward pass, while the unchanged AR weights check the draft. The output follows exactly the AR model's distribution. Uno is faster than leading speculative decoding methods at every batch size we tested, and can be built on top of existing open-weight models such as Qwen3-8B.
6 November 2026
Adversarial Concept Search: Predicting Compositional Errors From Feature Geometry
By: Jennifer Meng Lu, Ruochen Zhang, Isabelle Lee, David Alvarez-Melis, Ellie Pavlick, Naomi Saphra
Adversarial Concept Search (ACS) identifies meaningful conceptual scenarios that are likely to induce model failure without evaluating the model on specific inputs that instantiate those…
29 October 2026
Parametric Neural Control Differentiates Top Neural Network Models of Primate Visual Cortex
Hundreds of distinct models predict visual neural responses equally well. Are they all equally good models of visual neural tuning?
By: Jacob Prince*, Binxu Wang*, Thomas Fel, Akshay Jagadeesh, Parisa Vaziri, George Alvarez, Margaret Livingstone, and Talia Konkle
Leading AI models appeared to predict brain activity equally well, until asked to synthesize images that precisely control neural firing. Presented to monkeys during neural…
13 July 2026
Training Language Models That Can Continue to Learn
By: Tessa Han, Sebastian Bordt, Hanlin Zhang, and Sham Kakade
The usefulness of a pretrained model depends not only on what it has learned, but also on its plasticity, i.e., its ability to keep learning. However, despite its importance for downstream adaptation, model plasticity remains understudied in pretraining. The authors find that using stronger weight decay during pretraining improves model plasticity. Contrary to common practice, minimizing pretraining validation loss does not necessarily produce the most plastic model, i.e., the model that performs best after further training. The authors also find potential explanations for why weight decay improves plasticity, including more structured internal representations, simpler attention patterns, and reduced overfitting.
23 June 2026
Jailbreak Scaling Laws for Large Language Models: Polynomial–Exponential Crossover
By: Indranil Halder, Annesya Banerjee and Cengiz Pehlevan
The authors report that adversarial prompt-injection attacks on large language models can amplify attack success rate from the slow polynomial growth observed without injection to exponential growth with the number of inference-time samples.
27 April 2026
Deep RL Needs Deep Behavior Analysis: Exploring Implicit Planning by Model-Free Agents in Open-Ended Environments
By: Riley Simmons-Edler, Ryan Badman and Kanaka Rajan
Model-free RL agents trained on simple tasks can be evaluated by whether or not they achieve their simple rewards, but complex open-ended environments demand richer analysis. ForageWorld introduces a naturalistic foraging benchmark where agents must explore, remember food patches, and survive across procedurally generated arenas. In our experiments, naive RNN agents develop structured exploration strategies, multi-factor patch memory, and implicit future planning. These behaviors mirror those documented in foraging insects and rodents and are recoverable only using neuroscience-inspired analysis tools. The approach generalizes across arena sizes, transfers readily to other recurrent architectures, and opens a path toward behavioral and neural transparency in increasingly complex AI systems.
16 March 2026
Energy-Based Fine-Tuning: Beyond Next-Token Prediction
By: Samy Jelassi*, Mujin Kwun*, Rosie Zhao*, Yuanzhi Li, Nicolo Fusi, Yilun Du, Sham M. Kakade and Carles Domingo-Enrich*
The authors introduce a Energy-Based Fine-Tuning (EBFT), a method that matches the long-range statistics of model generations to ground-truth sequences in high-dimensional feature spaces. EBFT corrects the error amplification caused by standard teacher-forced next token training while improving downstream task performance. Experiments show that EBFT matches the accuracy improvements of RLVR while requiring no external reward signal or verifier. In contrast to RLVR, EBFT simultaneously improves the validation cross-entropy.
9 March 2026
InputDSA: Demixing then comparing recurrent and externally driven dynamics in complex systems
By: Ann Huang and Kanaka Rajan
The authors introduce InputDSA, a new method to measure the similarity between two complex systems when they are driven by external inputs, like biological neural circuits or reinforcement learning agents. The method disentangles each systems’ intrinsic dynamics from its input-driven effects, enabling highly accurate, robust, and efficient comparisons of those components.
4 March 2026
Structure, Disorder, and Dynamics in Task-Trained Recurrent Neural Circuits
By: David Clark,* Blake Bordelon,* Jacob Zavatone-Veth,* Cengiz Pehlevan
The authors develop a mean-field theory of task-trained recurrent networks that continuously interpolates between these regimes, and find evidence that macaque motor cortex is best captured by an intermediate level of task-specific recurrent restructuring.