AI
Paragraph Boundaries Are Not White Space:Compression Depth as the Signature of Hierarchical Structure
Publication: Hugging Face Daily Papers
Author: Not available in collected metadata
Source published: 19/09/2026, 20:00:00 UTC
Briefing created: 28/09/2026, 10:20:43 UTC
Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 28/09/2026, 10:18:24 UTC.
Briefings
This study investigates whether hierarchical textual structure in text can be identified by compression depth rather than positional encoding. Using hierarchical rotary positional encoding (hRoPE), the authors measured compression across eight corpus quantities in three constructs. They found that while embedding-based coherence was the closest match to cross-corpus ordering, no single quantity replicated the true signature of genuine paragraph structure. The key finding is that compression depth, not its location, serves as the reproducible signature of hierarchical structure.
Key points
- The study uses hierarchical rotary positional encoding (hRoPE) to represent paragraph, sentence, and token indices as separate channels.
- Compression depth was found to be the reproducible signature of genuine paragraph structure, distinguishing it from random labels or shallow compression.
Context and limitations
The evidence is limited to eight corpus quantities across three constructs, none fully reproducing cross-corpus ordering. The study does not provide causal claims or author attribution.
Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 28/09/2026, 10:20:43 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.
View archived briefing on IPFS
bafybeigqv5yuuyky2lodlsu4pew3skqdbg7zqbdok4q7mpbx2zi6g4ofju