Scalable oversight, the process by which weaker AI systems supervise stronger ones, has been proposed as a key strategy to control future superintelligent systems. However, it is still unclear how scalable oversight itself scales. To address this gap, we propose a framework that quantifies the probability of successful oversight as a function of the capabilities of the overseer and the system being overseen. Specifically, our framework models oversight as a game between capability-mismatched players; the players have oversight-specific Elo scores that are a piecewise-linear function of their general intelligence, with two plateaus corresponding to task incompetence and task saturation. We validate our framework with a modified version of the game Nim and then apply it to four oversight games: Mafia, Debate, Backdoor Code and Wargames. For each game, we find scaling laws that approximate how domain performance depends on general AI system capability. We then build on our findings in a theoretical study of Nested Scalable Oversight (NSO), a process in which trusted models oversee untrusted stronger models, which then become the trusted models in the next step. We identify conditions under which NSO succeeds and derive numerically (and in some cases analytically) the optimal number of oversight levels to maximize the probability of oversight success. We also apply our theory to our four oversight games, where we find that NSO success rates at a general Elo gap of 400 are 13.5% for Mafia, 51.7% for Debate, 10.0% for Backdoor Code, and 9.4% for Wargames; these rates decline further when overseeing stronger systems.

Abstract-- Autonomous Surface V ehicles (ASVs) play a crucial role in maritime operations, yet their navigation in shallow-water environments remains challenging due to dynamic disturbances and depth constraints. In this paper, we propose a reinforcement learning (RL) framework for ASV navigation under depth constraints, where the vehicle must reach a target while avoiding unsafe areas with only a single depth measurement per timestep from a downward-facing Single Beam Echosounder (SBES). T o enhance environmental awareness, we integrate Gaussian Process (GP) regression into the RL framework, enabling the agent to progressively estimate a bathymetric depth map from sparse sonar readings. This approach improves decision-making by providing a richer representation of the environment. Furthermore, we demonstrate effective sim-to-real transfer, ensuring that trained policies generalize well to real-world aquatic conditions. Experimental results validate our method's capability to improve ASV navigation performance while maintaining safety in challenging shallow-water environments. I. INTRODUCTION Autonomous Surface V ehicles (ASVs) are unmanned vessels increasingly employed for a variety of maritime operations, including environmental monitoring, search-and-rescue, and surveillance.

Today's robots attempt to learn new tasks by imitating human examples. These robots watch the human complete the task, and then try to match the actions taken by the human expert. However, this standard approach to visual imitation learning is fundamentally limited: the robot observes what the human does, but not why the human chooses those behaviors. Without understanding which features of the system or environment factor into the human's decisions, robot learners often misinterpret the human's examples. In practice, this results in causal confusion, inefficient learning, and robot policies that fail when the environment changes. We therefore propose a shift in perspective: instead of asking human teachers just to show what actions the robot should take, we also enable humans to intuitively indicate why they made those decisions. Under our paradigm human teachers attach markers to task-relevant objects and use natural language prompts to describe their state representation. Our proposed algorithm, CIVIL, leverages this augmented demonstration data to filter the robot's visual observations and extract a feature representation that aligns with the human teacher. CIVIL then applies these causal features to train a transformer-based policy that -- when tested on the robot -- is able to emulate human behaviors without being confused by visual distractors or irrelevant items. Our simulations and real-world experiments demonstrate that robots trained with CIVIL learn both what actions to take and why to take those actions, resulting in better performance than state-of-the-art baselines. From the human's perspective, our user study reveals that this new training paradigm actually reduces the total time required for the robot to learn the task, and also improves the robot's performance in previously unseen scenarios. See videos at our project website: https://civil2025.github.io

AI code generation tools have gained significant popularity among developers, who use them to assist in software development due to their capability to generate code. Existing studies mainly explored the quality, e.g., correctness and security, of AI-generated code, while in real-world software development, the prerequisite is to distinguish AI-generated code from human-written code, which emphasizes the need to explicitly declare AI-generated code by developers. To this end, this study intends to understand the ways developers use to self-declare AI-generated code and explore the reasons why developers choose to self-declare or not. We conducted a mixed-methods study consisting of two phases. In the first phase, we mined GitHub repositories and collected 613 instances of AI-generated code snippets. In the second phase, we conducted a follow-up practitioners' survey, which received 111 valid responses. Our research revealed the practices followed by developers to self-declare AI-generated code. Most practitioners (76.6%) always or sometimes self-declare AI-generated code. In contrast, other practitioners (23.4%) noted that they never self-declare AI-generated code. The reasons for self-declaring AI-generated code include the need to track and monitor the code for future review and debugging, and ethical considerations. The reasons for not self-declaring AI-generated code include extensive modifications to AI-generated code and the developers' perception that self-declaration is an unnecessary activity. We finally provided guidelines for practitioners to self-declare AI-generated code, addressing ethical and code quality concerns.

Event cameras offer a promising avenue for multi-view stereo depth estimation and Simultaneous Localization And Mapping (SLAM) due to their ability to detect blur-free 3D edges at high-speed and over broad illumination conditions. However, traditional deep learning frameworks designed for conventional cameras struggle with the asynchronous, stream-like nature of event data, as their architectures are optimized for discrete, image-like inputs. We propose a scalable, flexible and adaptable framework for pixel-wise depth estimation with event cameras in both monocular and stereo setups. The 3D scene structure is encoded into disparity space images (DSIs), representing spatial densities of rays obtained by back-projecting events into space via known camera poses. Our neural network processes local subregions of the DSIs combining 3D convolutions and a recurrent structure to recognize valuable patterns for depth prediction. Local processing enables fast inference with full parallelization and ensures constant ultra-low model complexity and memory costs, regardless of camera resolution. Experiments on standard benchmarks (MVSEC and DSEC datasets) demonstrate unprecedented effectiveness: (i) using purely monocular data, our method achieves comparable results to existing stereo methods; (ii) when applied to stereo data, it strongly outperforms all state-of-the-art (SOTA) approaches, reducing the mean absolute error by at least 42%; (iii) our method also allows for increases in depth completeness by more than 3-fold while still yielding a reduction in median absolute error of at least 30%. Given its remarkable performance and effective processing of event-data, our framework holds strong potential to become a standard approach for using deep learning for event-based depth estimation and SLAM. Project page: https://github.com/tub-rip/DERD-Net

Large Language Models (LLMs) have emerged as personalized assistants for users across a wide range of tasks -- from offering writing support to delivering tailored recommendations or consultations. Over time, the interaction history between a user and an LLM can provide extensive information about an individual's traits and preferences. However, open questions remain on how well LLMs today can effectively leverage such history to (1) internalize the user's inherent traits and preferences, (2) track how the user profiling and preferences evolve over time, and (3) generate personalized responses accordingly in new scenarios. In this work, we introduce the PERSONAMEM benchmark. PERSONAMEM features curated user profiles with over 180 simulated user-LLM interaction histories, each containing up to 60 sessions of multi-turn conversations across 15 real-world tasks that require personalization. Given an in-situ user query, i.e. query issued by the user from the first-person perspective, we evaluate LLM chatbots' ability to identify the most suitable response according to the current state of the user's profile. We observe that current LLMs still struggle to recognize the dynamic evolution in users' profiles over time through direct prompting approaches. As a consequence, LLMs often fail to deliver responses that align with users' current situations and preferences, with frontier models such as GPT-4.1, o4-mini, GPT-4.5, o1, or Gemini-2.0 achieving only around 50% overall accuracy, suggesting room for improvement. We hope that PERSONAMEM, along with the user profile and conversation simulation pipeline, can facilitate future research in the development of truly user-aware chatbots. Code and data are available at github.com/bowen-upenn/PersonaMem.

Abstract-- The field of robotic manipulation has advanced significantly in recent years. At the sensing level, several novel tactile sensors have been developed, capable of providing accurate contact information. On a methodological level, learning from demonstrations has proven an efficient paradigm to obtain performant robotic manipulation policies. The combination of both holds the promise to extract crucial contact-related information from the demonstration data and actively exploit it during policy rollouts. However, this integration has so far been underexplored, most notably in dynamic, contact-rich manipulation tasks where precision and reactivity are essential. This work therefore proposes a multimodal, visuotactile imitation learning framework that integrates a modular transformer architecture with a flow-based generative model, enabling efficient learning of fast and dexterous manipulation policies. We evaluate our framework on the dynamic, contact-rich task of robotic match lighting - a task in which tactile feedback influences human manipulation performance. The experimental results highlight the effectiveness of our approach and show that adding tactile information improves policy performance, thereby underlining their combined potential for learning dynamic manipulation from few demonstrations. I. INTRODUCTION Robotic manipulation still remains far from matching human dexterity and efficiency [1], [2]. A promising direction toward closing this gap is leveraging human demonstration data for learning robotic manipulation through imitation [3], [4], [5], thereby actively exploiting humans' task understanding and advanced manipulation capabilities. Although numerous studies have shown that access to touch sensing benefits human manipulation performance [6], [7], [8], the majority of current works in imitation learning for manipulation are still missing out on this modality [4], [5], [9].

Frontier models that generate extended reasoning traces inadvertently produce rich token sequences that can facilitate model distillation. Recognizing this vulnerability, model owners may seek sampling strategies that limit the effectiveness of distillation without compromising model performance. Antidistillation sampling provides exactly this capability. By strategically modifying a model's next-token probability distribution, antidistillation sampling poisons reasoning traces, rendering them significantly less effective for distillation while preserving the model's practical utility. For further details, see https://antidistillation.com.

Estimating the Kullback--Leibler (KL) divergence between language models has many applications, e.g., reinforcement learning from human feedback (RLHF), interpretability, and knowledge distillation. However, computing the exact KL divergence between two arbitrary language models is intractable. Thus, practitioners often resort to sampling-based estimators. While it is easy to fashion a simple Monte Carlo (MC) estimator that provides an unbiased estimate of the KL divergence between language models, this estimator notoriously suffers from high variance and can even result in a negative estimate of the KL divergence, a non-negative quantity. In this paper, we introduce a Rao--Blackwellized estimator that is unbiased and provably has variance less than or equal to that of the standard Monte Carlo estimator. In an empirical study on sentiment-controlled fine-tuning, we show that our estimator provides more stable KL estimates and reduces variance substantially. Additionally, we derive an analogous Rao--Blackwellized estimator of the gradient of the KL divergence, which leads to more stable training and produces models that more frequently appear on the Pareto frontier of reward vs. KL compared to the ones trained with the MC estimator of the gradient.

Accurate Subseasonal-to-Seasonal (S2S) forecasting is vital for decision-making in agriculture, energy production, and emergency management. However, it remains a challenging and underexplored problem due to the chaotic nature of the weather system. Recent data-driven studies have shown promising results, but their performance is limited by the inadequate incorporation of climate states and a model tendency to degrade, progressively losing fine-scale details and yielding over-smoothed forecasts. To overcome these limitations, we propose TianQuan-S2S, a global S2S forecasting model that integrates initial weather states with climatological means via incorporating climatology into patch embedding and enhancing variability capture through an uncertainty-augmented Transformer. Extensive experiments on the Earth Reanalysis 5 (ERA5) reanalysis dataset demonstrate that our model yields a significant improvement in both deterministic and ensemble forecasting over the climatology mean, traditional numerical methods, and data-driven models. Ablation studies empirically show the effectiveness of our model designs. Remarkably, our model outperforms skillful numerical ECMWF-S2S and advanced data-driven Fuxi-S2S in key meteorological variables.