Common methods for aligning already-capable models with desired behavior rely on the ability of humans to provide supervision. However, future superhuman models will surpass the capability of humans. Therefore, humans will only be able to weakly supervise superhuman models. This expected deficiency of human evaluation would weaken the safety of future AI systems. Scalable oversight and weak-to-strong generalization are two complementary approaches to tackle this issue. In this paper, we attempt to combine the strengths of these two approaches to further improve alignment. Specifically, we investigate ways of improving human supervision with a strong pretrained model and then supervise the strong model with enhanced weak human supervision. To make iterative empirical progress, we consider an analogy: can we use a strong model to improve weak model supervision and then use it to supervise the strong model? We empirically test it by finetuning a small weak model on ground truth labels with the additional help from a large strong model, and then finetuning the strong model on labels generated by the weak model. We find that debate can assist a weak model in extracting trustworthy information from an untrustworthy strong model, which provides leverage as context on samples when training a weak model. We also show that an ensemble of weak models helps exploit long arguments generated by strong model debaters and obtain a more robust supervision estimate. Extensive experiments on the OpenAI weak-to-strong NLP benchmarks show that the combination approach leads to better alignment, which indicates that debate has the potential to help weak-to-strong generalization.

The proliferation of web agents necessitates advanced navigation and interaction strategies within complex web environments. Current models often struggle with efficient navigation and action execution due to limited visibility and understanding of web structures. Our proposed R2D2 framework addresses these challenges by integrating two paradigms: Remember and Reflect. The Remember paradigm utilizes a replay buffer that aids agents in reconstructing the web environment dynamically, thus enabling the formulation of a detailed ``map'' of previously visited pages. This helps in reducing navigational errors and optimizing the decision-making process during web interactions. Conversely, the Reflect paradigm allows agents to learn from past mistakes by providing a mechanism for error analysis and strategy refinement, enhancing overall task performance. We evaluate R2D2 using the WEBARENA benchmark, demonstrating significant improvements over existing methods, including a 50% reduction in navigation errors and a threefold increase in task completion rates. Our findings suggest that a combination of memory-enhanced navigation and reflective learning promisingly advances the capabilities of web agents, potentially benefiting various applications such as automated customer service and personal digital assistants.

We introduce ADEPT: Adaptive Data ExPloiTation, a simple yet powerful framework to enhance the **data efficiency** and **generalization** in deep reinforcement learning (RL). Specifically, ADEPT adaptively manages the use of sampled data across different learning stages via multi-armed bandit (MAB) algorithms, optimizing data utilization while mitigating overfitting. Moreover, ADEPT can significantly reduce the computational overhead and accelerate a wide range of RL algorithms. We test ADEPT on benchmarks including Procgen, MiniGrid, and PyBullet. Extensive simulation demonstrates that ADEPT can achieve superior performance with remarkable computational efficiency, offering a practical solution to data-efficient RL. Our code is available at https://github.com/yuanmingqi/ADEPT.

This study addresses the pressing challenge of educational inclusion for students with special needs by proposing and developing an inclusive educational platform. Integrating machine learning, natural language processing, and cross-platform interfaces, the platform features key functionalities such as speech recognition functionality to support voice commands and text generation via voice input; real-time object recognition using the YOLOv5 model, adapted for educational environments; Grapheme-to-Phoneme (G2P) conversion for Text-to-Speech systems using seq2seq models with attention, ensuring natural and fluent voice synthesis; and the development of a cross-platform mobile application in Flutter with on-device inference execution using TensorFlow Lite. The results demonstrated high accuracy, usability, and positive impact in educational scenarios, validating the proposal as an effective tool for educational inclusion. This project underscores the importance of open and accessible technologies in promoting inclusive and quality education.

High-resolution land cover mapping plays a crucial role in addressing a wide range of global challenges, including urban planning, environmental monitoring, disaster response, and sustainable development. However, creating accurate, large-scale land cover datasets remains a significant challenge due to the inherent complexities of geospatial data, such as diverse terrain, varying sensor modalities, and atmospheric conditions. Synthetic Aperture Radar (SAR) imagery, with its ability to penetrate clouds and capture data in all-weather, day-and-night conditions, offers unique advantages for land cover mapping. Despite these strengths, the lack of benchmark datasets tailored for SAR imagery has limited the development of robust models specifically designed for this data modality. To bridge this gap and facilitate advancements in SAR-based geospatial analysis, we introduce OpenEarthMap-SAR, a benchmark SAR dataset, for global high-resolution land cover mapping. OpenEarthMap-SAR consists of 1.5 million segments of 5033 aerial and satellite images with the size of 1024$\times$1024 pixels, covering 35 regions from Japan, France, and the USA, with partially manually annotated and fully pseudo 8-class land cover labels at a ground sampling distance of 0.15--0.5 m. We evaluated the performance of state-of-the-art methods for semantic segmentation and present challenging problem settings suitable for further technical development. The dataset also serves the official dataset for IEEE GRSS Data Fusion Contest Track I. The dataset has been made publicly available at https://zenodo.org/records/14622048.

From my very personal point of view, the lifelong learning paradigm is a vague concept and is sometimes evoked for studying different scenarios that can be named otherwise (like transfer learning). However, I think that the framework studied here (from Balcan et al., 2015) is a very pertinent "lifelong problem". The authors present an honest work in the right direction. The proofs are not trivial but, as I explain below, the contribution appears to me insufficient for NIPS. The risk bound minimized by the learning algorithm may be very high, as it relies on the VC-dimension of the predictors.

Consider mentioning what sort of complexity you are after.

Figure 1: Training models to optimize test-time compute and learn "how to discover" correct responses, as opposed to the traditional learning paradigm of learning "what answer" to output. The major strategy to improve large language models (LLMs) thus far has been to use more and more high-quality data for supervised fine-tuning (SFT) or reinforcement learning (RL). Unfortunately, it seems this form of scaling will soon hit a wall, with the scaling laws for pre-training plateauing, and with reports that high-quality text data for training maybe exhausted by 2028, particularly for more difficult tasks, like solving reasoning problems which seems to require scaling current data by about 100x to see any significant improvement. The current performance of LLMs on problems from these hard tasks remains underwhelming (see example). There is thus a pressing need for data-efficient methods for training LLMs that extend beyond data scaling and can address more complex challenges.

The rebuttal answered all my technical questions, in particular the one about probability and expectations in Theorems 1(ii) and 3. My scores (which were given assuming the technicalities can be fixed) remain the same. The theoretical results do not feel that exciting. The regret bounds seem more or less what one might expect, and there is no comparison to bounds for other algorithms. I understand that there are no directly comparable bounds since you have both budget size and dimensionality of random projections as significant parameters, but still some discussion would be nice. The dependence on budget B in your bounds could perhaps be pointed out more explicitly.

Overall the paper presents a novel application of CNNs to co-evolution Protein Contact Prediction (PCP) with state-of-the-art results. The paper is a strong contribution to the field of coevolution PCP. The method has the potential to improve ab initio structure prediction accuracy, which could have wide ranging impacts on studying proteins without solved structural complexes. The paper itself contains flaws in the experimental setup, how the methods are described, and how the results are presented. Key issues: - Experimental setup: not enough unique training examples are used to compare to existing methods. More than 100K examples exist for this specific task, but only 307 are used.