Words in Motion: Extracting Interpretable Control Vectors for Motion Transformers

• URL: http://arxiv.org/abs/2406.11624v4
• Date: Thu, 20 Mar 2025 12:06:17 GMT
• Title: Words in Motion: Extracting Interpretable Control Vectors for Motion Transformers
• Authors: Omer Sahin Tas, Royden Wagner,
• Abstract summary: Transformer-based models generate hidden states that are difficult to interpret.<n>We analyze hidden states and modify them at inference, with a focus on motion forecasting.
• Score: 1.1510009152620668
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Transformer-based models generate hidden states that are difficult to interpret. In this work, we analyze hidden states and modify them at inference, with a focus on motion forecasting. We use linear probing to analyze whether interpretable features are embedded in hidden states. Our experiments reveal high probing accuracy, indicating latent space regularities with functionally important directions. Building on this, we use the directions between hidden states with opposing features to fit control vectors. At inference, we add our control vectors to hidden states and evaluate their impact on predictions. Remarkably, such modifications preserve the feasibility of predictions. We further refine our control vectors using sparse autoencoders (SAEs). This leads to more linear changes in predictions when scaling control vectors. Our approach enables mechanistic interpretation as well as zero-shot generalization to unseen dataset characteristics with negligible computational overhead.

Related papers

• Unified Human Localization and Trajectory Prediction with Monocular Vision [64.19384064365431]
  MonoTransmotion is a Transformer-based framework that uses only a monocular camera to jointly solve localization and prediction tasks. We show that by jointly training both tasks with our unified framework, our method is more robust in real-world scenarios made of noisy inputs.
  arXiv  Detail & Related papers  (2025-03-05T14:18:39Z)
• Activation Scaling for Steering and Interpreting Language Models [55.59689963561315]
  We argue that successfully intervening on a model is a prerequisite for interpreting its internal workings. We establish a three-term objective: a successful intervention should flip the correct with the wrong token and vice versa. Using gradient-based optimization, this objective lets us learn (and later evaluate) a specific kind of efficient and interpretable intervention.
  arXiv  Detail & Related papers  (2024-10-07T12:01:32Z)
• OOSTraj: Out-of-Sight Trajectory Prediction With Vision-Positioning Denoising [49.86409475232849]
  Trajectory prediction is fundamental in computer vision and autonomous driving. Existing approaches in this field often assume precise and complete observational data. We present a novel method for out-of-sight trajectory prediction that leverages a vision-positioning technique.
  arXiv  Detail & Related papers  (2024-04-02T18:30:29Z)
• Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Models [55.19497659895122]
  We introduce methods for discovering and applying sparse feature circuits. These are causally implicatedworks of human-interpretable features for explaining language model behaviors.
  arXiv  Detail & Related papers  (2024-03-28T17:56:07Z)
• Humanoid Locomotion as Next Token Prediction [84.21335675130021]
  Our model is a causal transformer trained via autoregressive prediction of sensorimotor trajectories. We show that our model enables a full-sized humanoid to walk in San Francisco zero-shot. Our model can transfer to the real world even when trained on only 27 hours of walking data, and can generalize commands not seen during training like walking backward.
  arXiv  Detail & Related papers  (2024-02-29T18:57:37Z)
• Social-Transmotion: Promptable Human Trajectory Prediction [65.80068316170613]
  Social-Transmotion is a generic Transformer-based model that exploits diverse and numerous visual cues to predict human behavior. Our approach is validated on multiple datasets, including JTA, JRDB, Pedestrians and Cyclists in Road Traffic, and ETH-UCY.
  arXiv  Detail & Related papers  (2023-12-26T18:56:49Z)
• LeTFuser: Light-weight End-to-end Transformer-Based Sensor Fusion for Autonomous Driving with Multi-Task Learning [16.241116794114525]
  We introduce LeTFuser, an algorithm for fusing multiple RGB-D camera representations. To perform perception and control tasks simultaneously, we utilize multi-task learning.
  arXiv  Detail & Related papers  (2023-10-19T20:09:08Z)
• MotionTrack: Learning Motion Predictor for Multiple Object Tracking [68.68339102749358]
  We introduce a novel motion-based tracker, MotionTrack, centered around a learnable motion predictor. Our experimental results demonstrate that MotionTrack yields state-of-the-art performance on datasets such as Dancetrack and SportsMOT.
  arXiv  Detail & Related papers  (2023-06-05T04:24:11Z)
• Backpack Language Models [108.65930795825416]
  We present Backpacks, a new neural architecture that marries strong modeling performance with an interface for interpretability and control. We find that, after training, sense vectors specialize, each encoding a different aspect of a word. We present simple algorithms that intervene on sense vectors to perform controllable text generation and debiasing.
  arXiv  Detail & Related papers  (2023-05-26T09:26:23Z)
• VISIT: Visualizing and Interpreting the Semantic Information Flow of Transformers [45.42482446288144]
  Recent advances in interpretability suggest we can project weights and hidden states of transformer-based language models to their vocabulary. We investigate LM attention heads and memory values, the vectors the models dynamically create and recall while processing a given input. We create a tool to visualize a forward pass of Generative Pre-trained Transformers (GPTs) as an interactive flow graph.
  arXiv  Detail & Related papers  (2023-05-22T19:04:56Z)
• STMT: A Spatial-Temporal Mesh Transformer for MoCap-Based Action Recognition [50.064502884594376]
  We study the problem of human action recognition using motion capture (MoCap) sequences. We propose a novel Spatial-Temporal Mesh Transformer (STMT) to directly model the mesh sequences. The proposed method achieves state-of-the-art performance compared to skeleton-based and point-cloud-based models.
  arXiv  Detail & Related papers  (2023-03-31T16:19:27Z)
• Safe Control Transitions: Machine Vision Based Observable Readiness Index and Data-Driven Takeover Time Prediction [2.799896314754614]
  We show that machine learning models which predict two metrics are robust to multiple camera views. We also introduce two metrics to evaluate the quality of control transitions following the takeover event.
  arXiv  Detail & Related papers  (2023-01-14T01:53:48Z)
• Extracting Latent Steering Vectors from Pretrained Language Models [14.77762401765532]
  We show that latent vectors can be extracted directly from language model decoders without fine-tuning. Experiments show that there exist steering vectors, which, when added to the hidden states of the language model, generate a target sentence nearly perfectly. We find that distances between steering vectors reflect sentence similarity when evaluated on a textual similarity benchmark.
  arXiv  Detail & Related papers  (2022-05-10T19:04:37Z)
• Control-Aware Prediction Objectives for Autonomous Driving [78.19515972466063]
  We present control-aware prediction objectives (CAPOs) to evaluate the downstream effect of predictions on control without requiring the planner be differentiable. We propose two types of importance weights that weight the predictive likelihood: one using an attention model between agents, and another based on control variation when exchanging predicted trajectories for ground truth trajectories.
  arXiv  Detail & Related papers  (2022-04-28T07:37:21Z)
• Stochastic Trajectory Prediction via Motion Indeterminacy Diffusion [88.45326906116165]
  We present a new framework to formulate the trajectory prediction task as a reverse process of motion indeterminacy diffusion (MID) We encode the history behavior information and the social interactions as a state embedding and devise a Transformer-based diffusion model to capture the temporal dependencies of trajectories. Experiments on the human trajectory prediction benchmarks including the Stanford Drone and ETH/UCY datasets demonstrate the superiority of our method.
  arXiv  Detail & Related papers  (2022-03-25T16:59:08Z)
• Trajectory Forecasting from Detection with Uncertainty-Aware Motion Encoding [121.66374635092097]
  Trajectories obtained from object detection and tracking are inevitably noisy. We propose a trajectory predictor directly based on detection results without relying on explicitly formed trajectories.
  arXiv  Detail & Related papers  (2022-02-03T09:09:56Z)
• Inverse reinforcement learning for autonomous navigation via differentiable semantic mapping and planning [20.66819092398541]
  This paper focuses on inverse reinforcement learning for autonomous navigation using distance and semantic category observations. We develop a map encoder, that infers semantic category probabilities from the observation sequence, and a cost encoder, defined as a deep neural network over the semantic features. We show that our approach learns to follow traffic rules in the autonomous driving CARLA simulator by relying on semantic observations of buildings, sidewalks, and road lanes.
  arXiv  Detail & Related papers  (2021-01-01T07:41:08Z)
• Positional Artefacts Propagate Through Masked Language Model Embeddings [16.97378491957158]
  We find cases of persistent outlier neurons within BERT and RoBERTa's hidden state vectors. We pre-train the RoBERTa-base models from scratch and find that the outliers disappear without using positional embeddings.
  arXiv  Detail & Related papers  (2020-11-09T12:49:39Z)
• Learning Navigation Costs from Demonstration with Semantic Observations [24.457042947946025]
  This paper focuses on inverse reinforcement learning (IRL) for autonomous robot navigation using semantic observations. We develop a map encoder, which infers semantic class probabilities from the observation sequence, and a cost encoder, defined as deep neural network over the semantic features. We show that our approach learns to follow traffic rules in the autonomous driving CARLA simulator by relying on semantic observations of cars, sidewalks and road lanes.
  arXiv  Detail & Related papers  (2020-06-09T04:35:57Z)
• Prediction and Generalisation over Directed Actions by Grid Cells [6.7141720056953895]
  Knowing how directed actions generalise to new situations is key to rapid generalisation. Recent work has proposed that neural grid codes provide an efficient representation of the state space. We show that a single set of eigenvectors can support predictions over arbitrary directed actions via action-specific eigenvalues.
  arXiv  Detail & Related papers  (2020-06-05T10:32:22Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.