Precomputing the Future-Offset Average in TriAttention
On the map: LLM reasoning → Test-time compute and efficient thinking
Abstract (arXiv)
TriAttention is a recent method for shrinking the KV cache of long-reasoning LLMs: it scores each cached key by how much attention it is likely to receive and evicts the lowest-scoring ones. Because a key does not know how far away its future queries will sit, the score is averaged over a ladder of 17 possible future distances. We point out that this average is free: the future distance enters the score only through the position-dependent rotation, so the whole 17-fold average collapses--exactly, by a one-line algebraic identity--into a single per-band weight that is computed once, offline. Scoring a key then costs one evaluation instead of seventeen, with no change to which keys get pruned. The saving is modest and lives entirely in TriAttention's pruning-score computation, not in the attention kernel; we present it as a small, exact complement to their method, and we confirm the identity numerically.
Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning. Builds on 3 of its 5 references that are in the map. First posted on arXiv 1 Jul 2026.
Builds on
3 of 5 references.
Closest ancestors · 1
Same direction, newest first
Other references · 2
- 2017Attention Is All You Needcited by 60,361
- 2021RoFormer: Enhanced Transformer with Rotary Position Embeddingcited by 4,703
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2025The Lottery LLM Hypothesis, Rethinking What Abilities Should LLM Compression Preserve?
- 2026Statistical Early Stopping for Reasoning Models
- 2025Cache What Lasts: Token Retention for Memory-Bounded KV Cache in LLMs
- 2024PRISM: Efficient Long-Range Reasoning With Short-Context LLMs
- 2026Taming Speculative Search for Test-Time Scaling in LLM Serving
- 2026Toward Skill-Native LLMs: Skill Entropy for Benchmarking and Training Long-Horizon Reasoning
- 2026LaTER: Efficient Test-Time Reasoning via Latent Exploration and Explicit Verification
- 2026Revisiting the Effectiveness of LLM Pruning for Test-Time Scaling