25 papers across AI, ML, NLP, and CV from the last 24 hours.
Today's batch reveals three converging pressures on the field. First, on-policy distillation and multi-teacher RL are maturing from lab curiosities into engineering disciplines — three papers independently grapple with the gap between token-level supervision and task-level objectives, diagnosing why dense teacher signals misalign with global constraints. Second, robotics is closing the gap between simulation and deployment: dexterous manipulation policies now transfer from sim to real via raw visuo-tactile perception, while Gaussian splatting enters the VLA policy stack as a zero-retraining fix for viewpoint shifts. Third, a cluster of papers on AI safety has shifted from adversarial attack/defense to structural monitoring — detecting covert coordination in latent agent states, backdoors via pre-NMS distribution shifts, and OoD hallucinations through elicited priors.
The standout is "Learned, Then Lost," which does the thing everyone talks about but nobody runs: two full pre-training runs differing by exactly one training example, repeated 24 times at 124M scale. It replaces estimation with measurement for a question that's been debated philosophically for years.
The batch collectively suggests that the field is pivoting from "can we build it?" to "can we govern what we built?" — precision, reliability, and operational discipline are displacing raw capability as the metrics that matter.
Zhu Zhang, Jixun Wang, Xiaoang Xu · 2026-08-19
On-policy distillation (OPD) trains a student on its own responses using dense token-level guidance from a stronger teacher. In long-context tasks, however, token-level teacher support can favor locally plausible responses that omit evidence distributed across the input or violate global task constraints. Task-specific verifiers, in contrast, evaluate task completion at the response level and may ...
Sotirios P. Chatzis, Loukas Papadoulas · 2026-08-19
Deep models for irregularly-sampled time series answer queries at arbitrary continuous timestamps, yet report nothing about how far each answer should be trusted. We show the attention layer itself can close that gap: with the right stochastic formulation, the pass that makes each prediction also reports, in closed form and at no extra cost, how far it should be trusted. We introduce Lévy Attentio...
Zachary Speck, Asa Shepard · 2026-08-19
A single training example's contribution to a finished model is normally estimated rather than measured, because measuring it takes two expensive full pre-training runs that differ in one row of one batch. We ran that counterfactual 24 times at a small scale. We trained 32 GPT-2 models at 124M parameters from scratch on OpenWebText, over four conditions and eight seeds. At step 200 of 9,536, at pe...
Zhenyao Cui, Siyuan Kan, Siyang Li · 2026-08-19
Accurate visual decoding can reveal how the brain represents visual information and recover perceived content from neural signals such as electroencephalography (EEG), with potential for neural communication. However, current EEG-to-image retrieval methods perform far below their within-subject counterparts for new users without labeled calibration, limiting real-world deployment. To understand th...
Zhenyao Cui, Siyuan Kan, Dingkun Liu · 2026-08-19
Decoding visual information from brain signals probes neural representations and enables neuro-rehabilitation and dream decoding. Recent brain-to-image retrieval approaches have achieved promising performance, typically by averaging many (up to 80) neural trials per image, requiring repeated stimulus presentation that increases latency, cost, and user burden. When only one or a few repetitions are...
Emanuele Luzio · 2026-08-19
A gradient-boosted ensemble predicts by summing one leaf value per tree. Read those values as coordinates rather than as intermediate results, and every instance becomes a point in R^M on which the model acts linearly: the score is the sum of the coordinates. This small change of view makes contrastive explanation exact. The difference between two instances is a vector that is identicall...
Boqiao Zhang, Godbless James, Sai Krishna Gottipati · 2026-08-19
Improving molecular properties, such as drug-likeness or binding affinity, is a recurring task in early-stage drug discovery. However, molecules optimized in an unconstrained chemical space have limited practical value if they cannot be synthesized. Policy Gradient for Forward Synthesis (PGFS) is a synthesis-aware reinforcement learning method for molecular improvement, but its use of reactant emb...
Huan-ang Gao, Haohan Chi, Yong Yan · 2026-08-19
Multi-teacher on-policy distillation (M-OPD) has emerged as a promising paradigm for consolidating domain-specialized reinforcement learning (RL) experts into a single generalist student via dense, token-level reward supervision. Despite its practical success, the optimization dynamics governing multi-teacher capability integration remain poorly understood, and open, rigorously reproducible recipe...
Ramneet Kaur, Pradyumna Chari, Ramesh Raskar · 2026-08-19
Language-model agents can communicate through continuous hidden states that are invisible in public transcripts, creating opportunities for covert harmful coordination. We introduce Verifiable Latent Alignments (VLA), an activation-aware framework for monitoring and steering these private communication channels. For every monitored decision, VLA links the private latent-state record and channel st...
George Andrikopoulos · 2026-08-19
Frontier language models are compared, marketed, and benchmarked on capability -- what their best or average output can achieve. I argue this measures the wrong axis. The models have saturated accuracy: their mean output lands on the target. What now separates one system from another in practice is precision: how tightly concentrated their outputs are around that target across repeated, identical ...
George Andrikopoulos · 2026-08-19
When an expert corrects an LLM assistant's error, the correction usually dies with the session, and the error class returns. I argue this is an operations problem, not a tooling problem: mechanisms for persisting corrections exist and are shipping, but the discipline for governing them -- versioning with provenance, recurrence monitoring, counter-metrics, retirement of stale rules -- does not. Wri...
Deep Kumar Ganguly, Jan Křetínský · 2026-08-19
An agent still learning its environment should be cautious while ignorant and bold once confident. The entropic value-at-risk captures this through a robust-optimization identity---a confidence level fixes the radius of a relative-entropy ball of alternative models---but that ball cannot reach catastrophes the nominal deems impossible, precisely what a safe agent must hedge. We instead use an opti...
Joy Jia Yin Lim, Xin Huang, Hao Peng · 2026-08-19
Large language model (LLM) agents can now post-train an LLM end-to-end. They can write code, launch training, evaluate checkpoints, and improve downstream performance, raising the prospect of AI-for-AI. We argue that this picture conflates two distinct capabilities: execution-level capability, iterating within a selected training strategy; and strategy-level capability, revising the high-level jud...
Longtian Wang, Zhengyu Zhao, Chenhao Lin · 2026-08-19
Object detection models deployed in safety-critical applications remain vulnerable to backdoor attacks that cause targeted misbehaviors when a hidden trigger is present. Existing detection methods either rely on trigger inversion or exploit architecture-specific assumptions, and critically, representative existing methods fail to generalize reliably to scene-level attacks, where a single trigger i...
Jihae Jeong, Junha Choi, Hwanjo Yu · 2026-08-19
Large vision-language models (LVLMs) often hallucinate, generating content that the input image does not support. Preventing such content during decoding calls for a candidate-specific measure of how strongly the image supports the token under consideration. The model's visual-token states offer a natural source of this evidence because projecting each state through the output head reveals which v...
Yechan Park, HyunJin Kim · 2026-08-19
This paper proposes a lightweight, plug-and-play framework that improves robustness to viewpoint shifts in Vision-Language-Action (VLA) policies without policy retraining. To our knowledge, this is the first approach to directly leverage 3D Gaussian-based novel-view synthesis for observation-space adaptation in VLA policies. Current VLA performance relies on the implicit assumption that training a...
Jack A. Johnson, Bartłomiej W. Papież · 2026-08-19
PSMA and FDG PET/CT visualise complementary biological information in prostate cancer. Combining both tracers could capture heterogeneous tumour phenotypes that may be missed by either alone, yet there is no consensus on effective deep learning architectures for fusing these modalities. We evaluated multimodal image-fusion strategies for automatic whole-body PET/CT lesion segmentation to estimate ...
Álvaro G. Iñesta, Mattia Ryffel, Amit H. Bermano · 2026-08-19
Music-driven character animation enables and enhances transformative applications in entertainment and interactive education. However, synthesizing realistic drumming motion from audio remains challenging due to the inherent tension between high-acceleration dynamics and the need for extreme spatial-temporal precision. Existing approaches, often reliant on motion matching or MIDI input, struggle w...
Jayjun Lee, Jessica Yin, Asif Rana · 2026-08-19
We introduce Accelerating Dexterity via Pre-Training (ADEPT), a large-scale reinforcement learning (RL) framework for learning sim-to-real transferable dexterity across high degree-of-freedom (DoF) robot embodiments that can solve long-horizon tasks directly from raw visuo-tactile perception. ADEPT pretrains a dexterous policy on a generic object reposing task, then post-trains downstream policies...
Ruiguo Zhong, Benshan Ma, Xiaolong Chen · 2026-08-19
Anticipating how scenes evolve under ego actions is fundamental to safe autonomous driving, yet the full potential of world models for decision-making remains unrealized. The critical challenge lies in ensuring that future modeling is not merely predictive, but decision-informative: the predicted future must directly shape which trajectory is selected. Existing approaches decouple future represent...
Yumin Lee, Hyoseok Ju, Giseop Kim · 2026-08-19
Long-term robot operation in evolving environments requires object-level understanding that persists across repeated revisits. Existing systems either overwrite history to maintain an up-to-date map or store semantic snapshots without consistent cross-session object identity, resulting in temporal amnesia: the systematic loss of object history that prevents answering queries such as "Where has the...
Bo Liu, Simon Yu, Yiding Jiang · 2026-08-19
Continuous self-improvement requires an ever-expanding pool of self-generated, diverse, adaptive goals. For language agents, existing training environment pools (hand-curated, statically synthesized, or frozen-verifier) keep the goal distribution fixed as the learner scales. We introduce SPADE (Self-Play in Adaptive Synthetic Executable Environments), a self-play RL framework in which a single LLM...
Steven Morse, Daniel Runfola, Trenton W. Ford · 2026-08-19
We present a novel application of embedding-based dynamic topic modeling techniques to detect and quantify topic drift at the comment level in a massive corpus. By leveraging pretrained language models to generate contextualized semantic embeddings for short text, we analyzed 12.7 billion Reddit comments spanning 2006 to 2022. Using unsupervised methods on these embeddings, we identify dynamically...
Yuga Iguchi, Paul Fearnhead · 2026-08-19
The empirical success of diffusion models in generative modelling has motivated theoretical work, including quantitative error bounds and qualitative analyses that characterise the different phases of denoising. We bring these two areas together by studying the adaptivity of diffusion models to the structured geometry of multimodal high-dimensional data that consists of multiple clusters in $\math...
Leo Schmidt-Traub, Frédéric Berdoz, Luca A. Lanzendörfer · 2026-08-19
Neural audio codecs based on Residual Vector Quantization (RVQ) have become the dominant discrete representation for token-based general audio generation, yet resynthesizing high-quality audio from coarse codec tokens remains an open problem and bounds the fidelity of every system that generates them. Prior work has framed resynthesis as a choice between discrete token prediction and continuous re...
This digest is generated automatically from arXiv submissions. Not affiliated with arXiv or Cornell University.