Skip to content
← Back to explorer

TriAttention: Efficient Long Reasoning with Trigonometric KV Compression

Weian Mao, Xi Lin, Wei Huang, Yuxin Xie, Tianfu Fu, Bohan Zhuang, Song Han, Yukang Chen · Apr 6, 2026 · Citations: 0

Automatic Metrics Law Pairwise Preference
Open arXiv Find Implementation RSS feed Shortlist (0)

Data freshness

Extraction: Fresh

Check recency before relying on this page for active eval decisions. Use stale pages as context and verify against current hub results.

Metadata refreshed

Apr 6, 2026, 5:58 PM

Recent

Extraction refreshed

Apr 10, 2026, 7:04 AM

Fresh

Extraction source

Persisted extraction

Confidence 0.70

Abstract

Extended reasoning in large language models (LLMs) creates severe KV cache memory bottlenecks. Leading KV cache compression methods estimate KV importance using attention scores from recent post-RoPE queries. However, queries rotate with position during RoPE, making representative queries very few, leading to poor top-key selection and unstable reasoning. To avoid this issue, we turn to the pre-RoPE space, where we observe that Q and K vectors are highly concentrated around fixed non-zero centers and remain stable across positions -- Q/K concentration. We show that this concentration causes queries to preferentially attend to keys at specific distances (e.g., nearest keys), with the centers determining which distances are preferred via a trigonometric series. Based on this, we propose TriAttention to estimate key importance by leveraging these centers. Via the trigonometric series, we use the distance preference characterized by these centers to score keys according to their positions, and also leverage Q/K norms as an additional signal for importance estimation. On AIME25 with 32K-token generation, TriAttention matches Full Attention reasoning accuracy while achieving 2.5x higher throughput or 10.7x KV memory reduction, whereas leading baselines achieve only about half the accuracy at the same efficiency. TriAttention enables OpenClaw deployment on a single consumer GPU, where long context would otherwise cause out-of-memory with Full Attention.

HFEPX Relevance Assessment

This paper has useful evaluation signal, but protocol completeness is partial; pair it with related papers before deciding implementation strategy.

Best use

Secondary protocol comparison source

Use if you need

A secondary eval reference to pair with stronger protocol papers.

Main weakness

No major weakness surfaced.

Trust level

Moderate

Eval-Fit Score

65/100 • Medium

Useful as a secondary reference; validate protocol details against neighboring papers.

Human Feedback Signal

Detected

Evaluation Signal

Detected

HFEPX Fit

Moderate-confidence candidate

Extraction confidence: Moderate

If you are doing eval pipeline work, start here:

Human Eval Hub LLM-as-Judge Hub Pairwise Preference Hub Tool-Use Eval Hub

Field Provenance & Confidence

Each key protocol field shows extraction state, confidence band, and data source so you can decide whether to trust it directly or validate from full text.

Human Feedback Types

strong

Pairwise Preference

Confidence: Moderate Source: Persisted extraction evidenced

Directly usable for protocol triage.

Evidence snippet: Extended reasoning in large language models (LLMs) creates severe KV cache memory bottlenecks.

Evaluation Modes

strong

Automatic Metrics

Confidence: Moderate Source: Persisted extraction evidenced

Includes extracted eval setup.

Evidence snippet: Extended reasoning in large language models (LLMs) creates severe KV cache memory bottlenecks.

Quality Controls

missing

Not reported

Confidence: Low Source: Persisted extraction missing

No explicit QC controls found.

Evidence snippet: Extended reasoning in large language models (LLMs) creates severe KV cache memory bottlenecks.

Benchmarks / Datasets

missing

Not extracted

Confidence: Low Source: Persisted extraction missing

No benchmark anchors detected.

Evidence snippet: Extended reasoning in large language models (LLMs) creates severe KV cache memory bottlenecks.

Reported Metrics

strong

Accuracy, Throughput

Confidence: Moderate Source: Persisted extraction evidenced

Useful for evaluation criteria comparison.

Evidence snippet: On AIME25 with 32K-token generation, TriAttention matches Full Attention reasoning accuracy while achieving 2.5x higher throughput or 10.7x KV memory reduction, whereas leading baselines achieve only about half the accuracy at the same efficiency.

Rater Population

missing

Unknown

Confidence: Low Source: Persisted extraction missing

Rater source not explicitly reported.

Evidence snippet: Extended reasoning in large language models (LLMs) creates severe KV cache memory bottlenecks.

Human Data Lens

  • Uses human feedback: Yes
  • Feedback types: Pairwise Preference
  • Rater population: Unknown
  • Unit of annotation: Unknown
  • Expertise required: Law
  • Extraction source: Persisted extraction

Evaluation Lens

  • Evaluation modes: Automatic Metrics
  • Agentic eval: None
  • Quality controls: Not reported
  • Confidence: 0.70
  • Flags: None

Protocol And Measurement Signals

Benchmarks / Datasets

No benchmark or dataset names were extracted from the available abstract.

Reported Metrics

accuracythroughput

Research Brief

Deterministic synthesis

We show that this concentration causes queries to preferentially attend to keys at specific distances (e.g., nearest keys), with the centers determining which distances are preferred via a trigonometric series. HFEPX signals include Pairwise Preference, Automatic Metrics with confidence 0.70. Updated from current HFEPX corpus.

Generated Apr 10, 2026, 7:04 AM · Grounded in abstract + metadata only

Key Takeaways

  • We show that this concentration causes queries to preferentially attend to keys at specific distances (e.g., nearest keys), with the centers determining which distances are…
  • Based on this, we propose TriAttention to estimate key importance by leveraging these centers.
  • Via the trigonometric series, we use the distance preference characterized by these centers to score keys according to their positions, and also leverage Q/K norms as an additional…

Researcher Actions

  • Compare its human-feedback setup against pairwise and rubric hubs.
  • Identify benchmark choices from full text before operationalizing conclusions.
  • Validate metric comparability (accuracy, throughput).

Caveats

  • Generated from title, abstract, and extracted metadata only; full-paper implementation details are not parsed.
  • Extraction confidence is probabilistic and should be validated for critical decisions.

Research Summary

Contribution Summary

  • We show that this concentration causes queries to preferentially attend to keys at specific distances (e.g., nearest keys), with the centers determining which distances are preferred via a trigonometric series.
  • Based on this, we propose TriAttention to estimate key importance by leveraging these centers.
  • Via the trigonometric series, we use the distance preference characterized by these centers to score keys according to their positions, and also leverage Q/K norms as an additional signal for importance estimation.

Why It Matters For Eval

  • Via the trigonometric series, we use the distance preference characterized by these centers to score keys according to their positions, and also leverage Q/K norms as an additional signal for importance estimation.

Researcher Checklist

  • Pass: Human feedback protocol is explicit

    Detected: Pairwise Preference

  • Pass: Evaluation mode is explicit

    Detected: Automatic Metrics

  • Gap: Quality control reporting appears

    No calibration/adjudication/IAA control explicitly detected.

  • Gap: Benchmark or dataset anchors are present

    No benchmark/dataset anchor extracted from abstract.

  • Pass: Metric reporting is present

    Detected: accuracy, throughput

Related Papers

Papers are ranked by protocol overlap, extraction signal alignment, and semantic proximity.

Need human evaluators for your AI research? Scale annotation with expert AI Trainers.

Post a Job Get a Quote