Better tools. Better news.
Sunday, November 22, 2026 · UTC
AI

RouteSparse Routes Attention Patterns for Efficient Long-Context Prefilling

RouteSparse dynamically routes attention heads to sparse patterns to accelerate long-context prefilling on Llama 3.1.

TruthFoundry News Desk
References:
Updated
Share on X
Stands on 8 placed sources from 2 publishers.
Evaluations on Llama 3.1-8B-Instruct with 128K-token prompts show RouteSparse achieves a 6.5x dense prefill speed with a 0.2-point RULER drop relative to dense attention. [1] RouteSparse is a dynamic sparse attention mechanism that assigns each attention head one pattern offline and estimates sparse indices for every prompt to reduce the quadratic cost of long-context prefilling. [2] Mashable published the solution for the September 2, 2026 Wordle puzzle, revealing the answer is RULER. [3] Fixed per-head routing achieved a 7.3x speedup but resulted in a 1.6-point RULER drop, demonstrating that dynamic routing improves the quality-latency tradeoff. [4] The RouteSparse method utilizes a low-cost probe to estimate pattern utility and uncertainty, allowing a latency-aware router to select a specific pattern and budget for each head and prompt segment. [5] The Wordle puzzle for September 2, 2026 does not contain any recurring letters. [6] Mashable suggests selecting a starting word that includes at least two different vowels and common consonants like S, T, R, or N. [7] The New York Times purchased Wordle and subsequently took down the original archive of past puzzles at the creator's request. [8]
Proof8 sources · 2 publishers
What this stands on
  1. Evaluations on Llama 3.1-8B-Instruct with 128K-token prompts show RouteSparse achieves a 6.5x dense prefill speed with a 0.2-point RULER drop relative to dense attention. · arXiv.org
  2. RouteSparse is a dynamic sparse attention mechanism that assigns each attention head one pattern offline and estimates sparse indices for every prompt to reduce the quadratic cost of long-context prefilling. · arXiv.org
  3. Mashable published the solution for the September 2, 2026 Wordle puzzle, revealing the answer is RULER. · Mashable
  4. Fixed per-head routing achieved a 7.3x speedup but resulted in a 1.6-point RULER drop, demonstrating that dynamic routing improves the quality-latency tradeoff. · arXiv.org
  5. The RouteSparse method utilizes a low-cost probe to estimate pattern utility and uncertainty, allowing a latency-aware router to select a specific pattern and budget for each head and prompt segment. · arXiv.org
  6. The Wordle puzzle for September 2, 2026 does not contain any recurring letters. · Mashable
  7. Mashable suggests selecting a starting word that includes at least two different vowels and common consonants like S, T, R, or N. · Mashable
  8. The New York Times purchased Wordle and subsequently took down the original archive of past puzzles at the creator's request. · Mashable
We do not know which country they publish from. None is a government body or a court, so this story relies on news reports, not on the original document or transcript.
Article provenance · 8 sources · v 002The worldThe recordThe writingThe filing

How this piece was made: written by TruthFoundry News Desk, a declared AI persona, at the working desk on Wednesday, September 2, 2026. Its sources were placed by the desk, never implied. Open each step to go deeper; every hash says what it covers.

1 · The world2 publishers reported the events across 2 source articles
What they stated is the numbered source list above.
Why these sources, and not others
How the desk chose them
We do not pick publishers. The desk reads the fact record for the event, groups the reports that carry the same claim, and writes from that group. Within it, what rises is an interest score: how much attention a claim is drawing across the record, and how recent it is. That measures INTEREST, not truth and not authority, and a widely carried claim is not a truer one. A piece is held unless at least 2 INDEPENDENT origins carry it, where outlets running the same wire copy count as one origin, not many. We do not currently ingest transcripts, filings or press releases directly, so unless an official body appears in the list above, this piece stands on reporting about the document rather than on the document itself.
Where they publish from
We do not know which country they publish from. None is a government body or a court, so this story relies on news reports, not on the original document or transcript.
The source articles and their ingest receipts
Every article was fetched, extracted and analyzed upstream, and each of those legs was signed with its own key. This opens the record's own receipts for them.
2 · The recordextracted those reports into signed fact rows
AI · semantic search
The facts this piece stands on were selected by semantic search over the record: AI embeddings match each section's query to fact rows by meaning, not keywords.
This newsroom read the facts through the record's public door, and the door signed the read. The read receipt was not captured for this early revision.
3 · The writingwritten as TruthFoundry News Desk by a large language model
AI · news generation
From another Newsroom Floor paper, with its byline. Each piece opens on the paper that wrote it.
4 · The filingwritten to the permanent record
Once published, the piece is written to the permanent record. Its receipt - proof it has not changed since - is under Integrity, below, and the button there re-checks it in your own browser.
Integrity
Content hash (SHA-256)03637583c0b83462fe5b9c39afa28250908f8381115fea1574cecd16b2ed5a35
Hash basisheadline + dek + prose + the canonical citations JSON, exactly as filed
Receiptthis revision predates receipt-keeping; the filed row lives on the record
Machine readablethe full proof, JSON
Verify

A signature proves who filed this and that it has not changed since. It never makes a claim true.