Pith. sign in

REVIEW 8 cited by

PoliFormer: Scaling On-Policy RL with Transformers Results in Masterful Navigators

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2406.20083 v1 pith:OAHEUKHQ submitted 2024-06-28 cs.RO cs.CV

classification cs.ROcs.CV
keywords navigationpoliformertrainedtransformeracrossmasterfulobjectresults
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

We present PoliFormer (Policy Transformer), an RGB-only indoor navigation agent trained end-to-end with reinforcement learning at scale that generalizes to the real-world without adaptation despite being trained purely in simulation. PoliFormer uses a foundational vision transformer encoder with a causal transformer decoder enabling long-term memory and reasoning. It is trained for hundreds of millions of interactions across diverse environments, leveraging parallelized, multi-machine rollouts for efficient training with high throughput. PoliFormer is a masterful navigator, producing state-of-the-art results across two distinct embodiments, the LoCoBot and Stretch RE-1 robots, and four navigation benchmarks. It breaks through the plateaus of previous work, achieving an unprecedented 85.5% success rate in object goal navigation on the CHORES-S benchmark, a 28.5% absolute improvement. PoliFormer can also be trivially extended to a variety of downstream applications such as object tracking, multi-object navigation, and open-vocabulary navigation with no finetuning.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 8 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. LangMap: A Human-Verified Benchmark for Hierarchical Open-Vocabulary Goal Navigation

    cs.CV 2026-02 conditional novelty 7.0 of 10

    LangMap is a human-verified navigation benchmark with 18K tasks spanning scene-, room-, region-, and instance-level goals in real-world 3D scans, covering 414 object categories.

  2. ABot-N1: Toward a General Visual Language Navigation Foundation Model

    cs.CV 2026-07 unverdicted novelty 6.0 of 10

    A slow–fast VLN model that routes five navigation tasks through CoT plus image-space pixel goals reaches SOTA on established and new urban benchmarks, including 77.3% POI arrival.

  3. MM-Nav: Multi-View VLA Model for Robust Visual Navigation via Multi-Expert Learning

    cs.RO 2025-10 conditional novelty 6.0 of 10

    A four-camera VLA navigation model trained by distilling multiple RL experts achieves strong simulation performance and qualitative real-world transfer.

  4. What Matters in RL-Based Methods for Object-Goal Navigation? An Empirical Study and A Unified Framework

    cs.RO 2025-10 conditional novelty 6.0 of 10

    In modular RL-based object-goal navigation, perception quality and test-time strategies dominate performance; policy architecture and observation-space choices contribute little under the tested settings.

  5. Move to Understand a 3D Scene: Bridging Visual Grounding and Exploration for Efficient and Versatile Embodied Navigation

    cs.CV 2025-07 conditional novelty 6.0 of 10

    MTU3D unifies visual grounding and frontier-based exploration in a single transformer, achieving state-of-the-art success rates on HM3D-OVON, GOAT-Bench, SG3D, and A-EQA after large-scale vision-language-exploration p...

  6. TrackVLA: Embodied Visual Tracking in the Wild

    cs.RO 2025-05 conditional novelty 6.0 of 10

    A single vision-language-action model jointly trained on recognition and tracking data follows described targets at the best reported levels on a public benchmark and transfers zero-shot from simulation to a real quad...

  7. Learning Category-level Last-meter Navigation from RGB Demonstrations of a Single-instance

    cs.RO 2025-12 conditional novelty 5.0 of 10

    An RGB-only imitation-learning policy trained on one chair generalizes to unseen chairs and environments for last-meter base positioning, though reported success depends on an added heuristic stopping rule and a 0.3 m...

  8. Spatially-Enhanced Recurrent Memory for Long-Range Mapless Navigation via End-to-End Reinforcement Learning

    cs.RO 2025-06 conditional novelty 5.0 of 10

    A modified recurrent unit with an input-multiplied gate improves spatial memory and long-range mapless navigation success rates by about 23.5% over standard RNNs in simulation and transfers zero-shot to a real robot.

Pith tools