Research map / Reinforcement learning / Visual and representation RL
Geometry of Uncertainty: Learning Metric Spaces for Multimodal State Estimation in RL
Alfredo Reichlin, Adriano Pacciarelli, Danica Kragic, Miguel Vasco
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use MetricMM fusion instead of ConCat under Failure corruption p=0.5 | SAC on Fetch Slide · Fetch Slide, Failure two modalities (p=0.5) | return: 4.01 -> 5.95 | Improved |
| use MetricMM instead of AMDF at low Patch corruption p=0.1 | SAC on Fetch PickAndPlace · Fetch PickAndPlace Patch (p=0.1) | return: 2.2 -> 1.91 | Hurt |
| use MetricMM instead of best baseline at Patch p=0.5 | SAC on Fetch PickAndPlace · Fetch PickAndPlace Patch (p=0.5) | return: -1.04 -> 1.43 | Improved |
| use MetricMM instead of baselines when only point cloud is corrupted | SAC on Fetch Slide · Fetch Slide Failure, point cloud only | return: 9.27 -> 7.87 | No clear effect |
| use MetricMM when single modality image-only corrupted (vs ConCat) | SAC on Fetch Slide · Fetch Slide Failure, image only | return: 0.28 -> 7.29 | Improved |
| inject noise during training of baselines (ablation) | CURL/GMC/ConCat/LinearComb on Hopper-v5 · Hopper-v5 | training return | Hurt |
| add multimodal invariance loss (ablation) | MetricMM · pendulum with Gaussian noise | return under noise | Improved |
| add metric loss pair L+ and L- (ablation) | MetricMM · pendulum with Gaussian noise | return under noise | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- CURL — CURL: Contrastive Unsupervised Representations for Reinforcement Learning (ICML 2020) · better on 6 of 6 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- METRA: Scalable Unsupervised RL with Metric-Aware Abstraction (ICLR 2024 · similarity 0.88)
- Temporal Distance-aware Transition Augmentation for Offline Model-based Reinforcement Learning (ICML 2025 · similarity 0.86)
- Offline RL with Observation Histories: Analyzing and Improving Sample Complexity (ICLR 2024 · similarity 0.86)
- Imagine Beyond ! Distributionally Robust Autoencoding for State Space Coverage in Online Reinforcement Learning (NeurIPS 2025 · similarity 0.86)
- Hitting Time Isomorphism for Multi-Stage Planning with Foundation Policies (NeurIPS 2026 · similarity 0.86)
- Learning from Reward-Free Offline Data: A Case for Planning with Latent Dynamics Models (NeurIPS 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
