人形控制

It has been reported that LLMs can recognize their own writing. As this has potential implications for AI safety, yet is relatively understudied, we investigate the phenomenon, seeking to establish: whether it robustly occurs at the behavioral level, how the observed behavior is achieved, and whether it can be controlled. First, we find that the Llama3-8b–Instruct chat model - but not the base Llama3-8b model - can reliably distinguish its own outputs from those of humans, and present evidence that the chat model is likely using its experience with its own outputs, acquired during post-training, to succeed at the writing recognition task. Second, we identify a vector in the residual stream of the model that is differentially activated when the model makes a correct self-written-text recognition judgment, show that the vector activates in response to information relevant to self-authorship, present evidence that the vector is related to the concept of ``self'' in the model, and demonstrate that the vector is causally related to the model’s ability to perceive and assert self-authorship. Finally, we show that the vector can be used to control both the model’s behavior and its perception, steering the model to claim or disclaim authorship by applying the vector to the model’s output as it generates it, and steering the model to believe or disbelieve it wrote arbitrary texts by applying the vector to them as the model reads them.

Reinforcement learning (RL) is rapidly reaching and surpassing human-level control capabilities. However, state-of-the-art RL algorithms often require timesteps and reaction times significantly faster than human capabilities, which is impractical in real-world settings and typically necessitates specialized hardware. We introduce Sequence Reinforcement Learning (SRL), an RL algorithm designed to produce a sequence of actions for a given input state, enabling effective control at lower decision frequencies. SRL addresses the challenges of learning action sequences by employing both a model and an actor-critic architecture operating at different temporal scales. We propose a "temporal recall" mechanism, where the critic uses the model to estimate intermediate states between primitive actions, providing a learning signal for each individual action within the sequence. Once training is complete, the actor can generate action sequences independently of the model, achieving model-free control at a slower frequency. We evaluate SRL on a suite of continuous control tasks, demonstrating that it achieves performance comparable to state-of-the-art algorithms while significantly reducing actor sample complexity. To better assess performance across varying decision frequencies, we introduce the Frequency-Averaged Score (FAS) metric. Our results show that SRL significantly outperforms traditional RL algorithms in terms of FAS, making it particularly suitable for applications requiring variable decision frequencies. Furthermore, we compare SRL with model-based online planning, showing that SRL achieves comparable FAS while leveraging the same model during training that online planners use for planning.

Fixed-size learned representations (dense representations, or embeddings) are widely used in many machine learning applications across language, vision or speech modalities. This paper investigates the role of the temperature parameter in contrastive training for text embeddings. We shed light on the impact this parameter has on the intrinsic dimensionality of the embedding spaces obtained, and show that lower intrinsic dimensionality is further correlated with effective compression of embeddings. We still observe a trade-off between absolute performance and effective compression and we propose temperature aggregation methods which reduce embedding size by an order of magnitude with minimal impact on quality.

We initiate a study of the geometry of the visual representation space ---the information channel from the vision encoder to the action decoder--- in an image-based control pipeline learned from behavior cloning. Inspired by the phenomenon of *neural collapse* (NC) in image classification, we empirically demonstrate the prevalent emergence of a similar *law of clustering* in the visual representation space. Specifically, - In discrete image-based control (e.g., Lunar Lander), the visual representations cluster according to the natural discrete action labels; - In continuous image-based control (e.g., Planar Pushing and Block Stacking), the clustering emerges according to ``control-oriented'' classes that are based on (a) the relative pose between the object and the target in the input or (b) the relative pose of the object induced by expert actions in the output. Each of the classes corresponds to one relative pose orthant (REPO). Beyond empirical observation, we show such a law of clustering can be leveraged as an algorithmic tool to improve test-time performance when training a policy with limited expert demonstrations. Particularly, we pretrain the vision encoder using NC as a regularization to encourage control-oriented clustering of the visual features. Surprisingly, such an NC-pretrained vision encoder, when finetuned end-to-end with the action decoder, boosts the test-time performance by 10% to 35%. Real-world vision-based planar pushing experiments confirmed the surprising advantage of control-oriented visual representation pretraining.

We introduce Enhancing Graph of Thoughts (EGoT), a method designed to enhance the performance of large language models (LLMs) on complex reasoning tasks. EGoT automates the process of generating accurate responses using given data and a base prompt. The process consists of several steps: It obtains an initial response from the answering node using the base prompt. Evaluation node evaluates the response and generates reasoning for it, utilizing the score's probabilities to enhance evaluation accuracy. The reasoning from both the answering node and the evaluation node is aggregated to identify the problem in the response. This aggregated reasoning is incorporated into the base prompt to obtain an enhanced response. These steps are organized in a graph architecture, where the final leaf nodes are merged to produce a final response. As the graph descends, the temperature is lowered using Cosine Annealing and scoring, to explore diverse responses with earlier nodes and to focus on precise responses with later nodes. The minimum temperature in Cosine Annealing is adjusted based on scoring, ensuring that nodes with low scores continue to explore diverse responses, while those with high scores confirm accurate responses. In sorting 256 elements using GPT-4o mini, EGoT performs 88.31\% accuracy, while GoT (Graph of Thoughts) achieves 84.37\% accuracy. In the frozen lake problem using GPT-4o, EGoT averages 0.55 jumps or falls into the hole, while ToT (Tree of Thoughts) averages 0.89.

The Koopman theory, which enables the transformation of nonlinear systems into linear representations, is a powerful and efficient tool to model and control nonlinear systems. However, the ability of the Koopman operator to model complex systems, particularly time-varying systems, is limited by the fixed linear state-space representation. To address the limitation, the large language model, Mamba, is considered a promising strategy for enhancing modeling capabilities while preserving the linear state-space structure. In this paper, we propose a new framework, the Mamba-based Koopman operator (MamKO), which provides enhanced model prediction capability and adaptability, as compared to Koopman models with constant Koopman operators. Inspired by the Mamba structure, MamKO generates Koopman operators from online data; this enables the model to effectively capture the dynamic behaviors of the nonlinear system over time. A model predictive control system is then developed based on the proposed MamKO model. The modeling and control performance of the proposed method is evaluated through experiments on benchmark time-invariant and time-varying systems. The experimental results demonstrate the superiority of the proposed approach. Additionally, we perform ablation experiments to test the effectiveness of individual components of MamKO. This approach unlocks new possibilities for integrating large language models with control frameworks, and it achieves a good balance between advanced modeling capabilities and real-time control implementation efficiency.

Policies developed through Reinforcement Learning (RL) and Imitation Learning (IL) have shown great potential in continuous control tasks, but real-world applications often require adapting trained policies to unforeseen requirements. While fine-tuning can address such needs, it typically requires additional data and access to the original training metrics and parameters. In contrast, an online planning algorithm, if capable of meeting the additional requirements, can eliminate the necessity for extensive training phases and customize the policy without knowledge of the original training scheme or task. In this work, we propose a generic online planning algorithm for customizing continuous-control policies at the execution time, which we call Residual-MPPI. It can customize a given prior policy on new performance metrics in few-shot and even zero-shot online settings, given access to the prior action distribution alone. Through our experiments, we demonstrate that the proposed Residual-MPPI algorithm can accomplish the few-shot/zero-shot online policy customization task effectively, including customizing the champion-level racing agent, Gran Turismo Sophy (GT Sophy) 1.0, in the challenging car racing scenario, Gran Turismo Sport (GTS) environment. Code for MuJoCo experiments is included in the supplementary and will be open-sourced upon acceptance. Demo videos are available on our website: https://sites.google.com/view/residual-mppi.

This work investigates the selection of high-quality pre-training data from massive corpora to enhance LMs' capabilities for downstream usage. We formulate data selection as a generalized Optimal Control problem, which can be solved theoretically by Pontryagin's Maximum Principle (PMP), yielding a set of necessary conditions that characterize the relationship between optimal data selection and LM training dynamics. Based on these theoretical results, we introduce **P**MP-based **D**ata **S**election (**PDS**), a framework that approximates optimal data selection by solving the PMP conditions. In our experiments, we adopt PDS to select data from CommmonCrawl and show that the PDS-selected corpus accelerates the learning of LMs and constantly boosts their performance on a wide range of downstream tasks across various model sizes. Moreover, the benefits of PDS extend to ~400B models trained on ~10T tokens, as evidenced by the extrapolation of the test loss curves according to the Scaling Laws. PDS also improves data utilization when the pre-training data is limited, by reducing the data demand by 1.8 times, which helps mitigate the quick exhaustion of available web-crawled corpora. Our code, model, and data can be found at https://github.com/microsoft/LMOps/tree/main/data_selection.

We propose Reference-Based Modulation (RB-Modulation), a new plug-and-play solution for training-free personalization of diffusion models. Existing training-free approaches exhibit difficulties in (a) style extraction from reference images in the absence of additional style or content text descriptions, (b) unwanted content leakage from reference style images, and (c) effective composition of style and content. RB-Modulation is built on a novel stochastic optimal controller where a style descriptor encodes the desired attributes through a terminal cost. The resulting drift not only overcomes the difficulties above, but also ensures high fidelity to the reference style and adheres to the given text prompt. We also introduce a cross-attention-based feature aggregation scheme that allows RB-Modulation to decouple content and style from the reference image. With theoretical justification and empirical evidence, our test-time optimization framework demonstrates precise extraction and control of *content* and *style* in a training-free manner. Further, our method allows a seamless composition of content and style, which marks a departure from the dependency on external adapters or ControlNets. See project page: https://rb-modulation.github.io/ for code and further details.

Reinforcement Learning (RL) has achieved remarkable success in various continuous control tasks, such as robot manipulation and locomotion. Different to mainstream RL which makes decisions at individual steps, recent studies have incorporated action repetition into RL, achieving enhanced action persistence with improved sample efficiency and superior performance. However, existing methods treat all action dimensions as a whole during repetition, ignoring variations among them. This constraint leads to inflexibility in decisions, which reduces policy agility with inferior effectiveness. In this work, we propose a novel repetition framework called SDAR, which implements Spatially Decoupled Action Repetition through performing closed-loop act-or-repeat selection for each action dimension individually. SDAR achieves more flexible repetition strategies, leading to an improved balance between action persistence and diversity. Compared to existing repetition frameworks, SDAR is more sample efficient with higher policy performance and reduced action fluctuation. Experiments are conducted on various continuous control scenarios, demonstrating the effectiveness of spatially decoupled repetition design proposed in this work.

The growing use of 3D point cloud data in autonomous vehicles (AVs) has raised serious privacy concerns, particularly due to the sensitive information that can be extracted from 3D data. While model inversion attacks have been widely studied in the context of 2D data, their application to 3D point clouds remains largely unexplored. To fill this gap, we present the first in-depth study of model inversion attacks aimed at restoring 3D point cloud scenes. Our analysis reveals the unique challenges, the inherent sparsity of 3D point clouds and the ambiguity between empty and non-empty voxels after voxelization, which are further exacerbated by the dispersion of non-empty voxels across feature extractor layers. To address these challenges, we introduce ConcreTizer, a simple yet effective model inversion attack designed specifically for voxel-based 3D point cloud data. ConcreTizer incorporates Voxel Occupancy Classification to distinguish between empty and non-empty voxels and Dispersion-Controlled Supervision to mitigate non-empty voxel dispersion. Extensive experiments on widely used 3D feature extractors and benchmark datasets, such as KITTI and Waymo, demonstrate that ConcreTizer concretely restores the original 3D point cloud scene from disrupted 3D feature data. Our findings highlight both the vulnerability of 3D data to inversion attacks and the urgent need for robust defense strategies.

When symmetry is present in the loss function, the model is likely to be trapped in a low-capacity state that is sometimes known as a ``collapse." Being trapped in these low-capacity states can be a major obstacle to training across many scenarios where deep learning technology is applied. We first prove two concrete mechanisms through which symmetries lead to reduced capacities and ignored features during training and inference. We then propose a simple and theoretically justified algorithm, \textit{syre}, to remove almost all symmetry-induced low-capacity states in neural networks. When this type of entrapment is especially a concern, removing symmetries with the proposed method is shown to correlate well with improved optimization or performance. A remarkable merit of the proposed method is that it is model-agnostic and does not require any knowledge of the symmetry.

Model-based reinforcement learning (MBRL) has shown promise for improving sample efficiency and decision-making in complex environments. However, existing methods face challenges in training stability, robustness to noise, and computational efficiency. In this paper, we propose Bisimulation Metric for Model Predictive Control (BS-MPC), a novel approach that incorporates bisimulation metric loss in its objective function to directly optimize the encoder. This optimization enables the learned encoder to extract intrinsic information from the original state space while discarding irrelevant details. BS-MPC improves training stability, robustness against input noise, and computational efficiency by reducing training time. We evaluate BS-MPC on both continuous control and image-based tasks from the DeepMind Control Suite, demonstrating superior performance and robustness compared to state-of-the-art baseline methods.

Video generation technologies are developing rapidly and have broad potential applications. Among these technologies, camera control is crucial for generating professional-quality videos that accurately meet user expectations. However, existing camera control methods still suffer from several limitations, including control precision and the neglect of the control for subject motion dynamics. In this work, we propose I2VControl-Camera, a novel camera control method that significantly enhances controllability while providing adjustability over the strength of subject motion. To improve control precision, we employ point trajectory in the camera coordinate system instead of only extrinsic matrix information as our control signal. To accurately control and adjust the strength of subject motion, we explicitly model the higher-order components of the video trajectory expansion, not merely the linear terms, and design an operator that effectively represents the motion strength. We use an adapter architecture that is independent of the base model structure. Experiments on static and dynamic scenes show that our framework outperformances previous methods both quantitatively and qualitatively. Project page: https://wanquanf.github.io/I2VControlCamera.

Unsupervised reinforcement learning (RL) aims at pre-training models that can solve a wide range of downstream tasks in complex environments. Despite recent advancements, existing approaches suffer from several limitations: they may require running an RL process on each task to achieve a satisfactory performance, they may need access to datasets with good coverage or well-curated task-specific samples, or they may pre-train policies with unsupervised losses that are poorly correlated with the downstream tasks of interest. In this paper, we introduce FB-CPR, which regularizes unsupervised zero-shot RL based on the forward-backward (FB) method towards imitating trajectories from unlabeled behaviors. The resulting models learn useful policies imitating the behaviors in the dataset, while retaining zero-shot generalization capabilities. We demonstrate the effectiveness of FB-CPR in a challenging humanoid control problem. Training FB-CPR online with observation-only motion capture datasets, we obtain the first humanoid behavioral foundation model that can be prompted to solve a variety of whole-body tasks, including motion tracking, goal reaching, and reward optimization. The resulting model is capable of expressing human-like behaviors and it achieves competitive performance with task-specific methods while outperforming state-of-the-art unsupervised RL and model-based baselines.

Controlled generation with pre-trained Diffusion and Flow Matching models has vast applications. One strategy for guiding ODE-based generative models is through optimizing a target loss $R(x_1)$ while staying close to the prior distribution. Along this line, some recent work showed the effectiveness of guiding flow model by differentiating through its ODE sampling process. Despite the superior performance, the theoretical understanding of this line of methods is still preliminary, leaving space for algorithm improvement. Moreover, existing methods predominately focus on Euclidean data manifold, and there is a compelling need for guided flow methods on complex geometries such as SO(3), which prevails in high-stake scientific applications like protein design. We present OC-Flow, a general and theoretically grounded training-free framework for guided flow matching using optimal control. Building upon advances in optimal control theory, we develop effective and practical algorithms for solving optimal control in guided ODE-based generation and provide a systematic theoretical analysis of the convergence guarantee in both Euclidean and SO(3). We show that existing backprop-through-ODE methods can be interpreted as special cases of Euclidean OC-Flow. OC-Flow achieved superior performance in extensive experiments on text-guided image manipulation, conditional molecule generation, and all-atom peptide design.