English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Pretrained representations from large-scale vision models have boosted theperformance of downstream embodied policy learning. We look to understandwhether additional self-supervised pretraining on exploration trajectories canbuild on these general-purpose visual representations to better supportembodied planning in realistic environments. We evaluated four common auxiliarylosses in embodied AI, two hindsight-based losses, and a standard imitationlearning loss, by pretraining the agent's visual compression module and statebelief representations with each objective and using CLIP as a representativevisual backbone. The learned representations are then frozen for downstreammulti-step evaluation on two goal-directed tasks. Surprisingly, we find thatimitation learning on these exploration trajectories out-performs all otherauxiliary losses even despite the exploration trajectories being dissimilarfrom the downstream tasks. This suggests that imitation of exploration may be''all you need'' for building powerful planning representations. Additionally,we find that popular auxiliary losses can benefit from simple modifications toimprove their support for downstream planning ability.

Understanding Representations Pretrained with Auxiliary Losses for Embodied Agent Planning