Integrate ICRA 2026 into cross-paper in-depth reviews
Wove ICRA 2026 developments into 9 cross-paper reviews (drawing only from the
already-verified ICRA topic/per-paper pages — no new web claims), mapping ICRA
papers onto each review's existing taxonomy:
- VLA-Architecture, Dexterous-Manipulation, RL, VLA-Memory, Goal-Image-Conditioning,
System-0-1-2, Cross-Embodiment, VLM-Action-Connection, WAM-vs-VLA-Robustness
Six of these had zero ICRA content before. 0 dangling wikilinks.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add ICRA 2026 (Vienna) VLA & manipulation analysis
- New venue: ICRA hub + ICRA-2026 index + ICRA-2026 VLA & Manipulation Survey
(5,088 submissions record; ~2,820 archival + 131 late-breaking; 11 themed clusters)
- 10 web-verified per-paper pages: VLA-Reasoner, FD-VLA, OmniVLA(nav), Dexora,
Flow Policy Optimization, MAP-VLA, Goal-VLA, LightVLA, VLA-Practicality, Galaxea+G0
- Cross-linked into Home, sidebar, and topic reviews (Architecture, RL, Memory,
Goal-Image-Conditioning, Dexterous, AsyncVLA)
- 0 dangling wikilinks across 212 pages
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Source-verification audit: fix fabrications, rename PI slugs, link hygiene, add foundational pages
- Re-verified all 195 pages vs original sources; removed 47 confirmed fabricated
tables/numbers, restored 19 false-positive deletions (full-PDF re-check)
- Renamed PI tech-report slugs ICLR-2026-pi07/pi06/RECAP -> PI-pi07/PI-pi06/PI-RECAP
(these are PI technical reports, not ICLR 2026 papers); updated 171 wikilinks
- Fixed 60 broken wikilinks -> 0 dangling across the wiki
- Added foundational pages: OpenVLA, ReKep, AgiBot World Colosseo, RoboBrain 2.0;
linked from Home + sidebar + venue indexes
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add goal-image-conditioning cross-paper review
Cross-paper review of 15 goal / subgoal / future-frame conditioned
VLAs from 2023-2026, anchored on how each feeds its goal image into
the low-level policy and compared axis-by-axis against pi0.7's
BAGEL-14B recipe.
Four design axes: source (who produces the goal), consumption (how
the policy ingests it), granularity (terminal / single subgoal /
H-frame video / multi-view), cadence (once / per-subtask / fixed step
/ receding horizon / async / inline).
Five mechanism clusters:
- A. Editing-diffusion subgoal: SuSIE, GR-MG
- B. Text-to-video + inverse dynamics: UniPi, HiP, VLP, RoboDreamer
- C. Human-video prior + cross-attention: Gen2Act
- D. Unified / co-generated frames and actions: CoT-VLA, dVLA,
Unified Diffusion VLA, VideoVLA, GR-1/2, Chain-of-Action
- E. Foundation world model as prompt: pi0.7 + BAGEL-14B
Per-paper cards, comparison table of pi0.7 vs. the field on 12
dimensions, mermaid dataflow contrast (pi0.7 vs. typical SuSIE-style
competitor), coupling-spectrum mermaid, open questions, decision
guide. Wired into Home.md cross-paper table and What's new, plus
_Sidebar.md.