Loading
Life sciences · Preprint
arXiv · September 4, 2026
Early or partial results. Treat as a signal, not a conclusion.
BeaconKV is a training-free KV cache compression algorithm that uses beacon queries—compact representatives of query clusters—to anticipate future token importance in long-horizon reasoning without storing full query history. The method reports up to 5.8× memory reduction and 4.3× throughput improvement across four open-source large reasoning models, but has not undergone peer review and lacks independent validation.
Preprint. Four open-source Large Reasoning Models evaluated on diverse reasoning benchmarks. Intervention: BeaconKV: training-free KV cache compression using beacon queries as compact query cluster representatives. Compared with: Existing KV cache compression methods (unspecified in abstract).
Queries corresponding to Thought Revisiting Tokens cluster into a small number of similarity groups in embedding space BeaconKV achieves up to 5.8× memory reduction while nearly preserving full cache accuracy Throughput improvement of over 4.3× reported across tested models
Safety was not reported in the material analysed. Check the source before drawing any conclusion about harm.
The source did not state who this applies to in practice.
A training-free algorithmic method for KV cache compression in reasoning models, demonstrated on benchmarks with reported memory and throughput improvements, but lacking peer review and clinical/clinical-adjacent validation.
As stated by the source record.
Quoted from the source exactly as published.
Graded across the dimensions that decide whether you should act, each from what the source actually supports. There is no single score, and where a dimension was not assessed it says so.
Large Reasoning Models (LRMs) achieve superior problem-solving through extended Chain-of-Thought (CoT) generation, but the resulting key-value (KV) cache grows linearly with sequence length and creates severe memory bottlenecks, often exceeding GPU capacity for long reasoning traces. Existing KV cache compression methods rely on recent queries to estimate future token importance, implicitly assuming these serve as reliable proxies for future attention patterns. We demonstrate that this assumption fails in long-horizon reasoning: certain decoding steps generate Thought Revisiting Tokens (TRT) that re-attend to distant previous context, such as task-solving plans formulated early in the trace. Through systematic analysis, we discover that queries corresponding to the TRT cluster into a small number of similarity groups in the embedding space. Based on this insight, we propose BeaconKV, a training-free KV cache compression method that maintains beacon queries, compact representatives for each global query cluster, to anticipate which KV pairs will be revisited without storing the entire query history. Across four open-source LRMs and diverse reasoning benchmarks, BeaconKV generally outperforms existing compression methods, achieving up to $5.8\times$ memory reduction while nearly preserving full cache accuracy and improving throughput by over $4.3\times$.
Taken from the source record, never inferred. Follow any of these and new work involving them reaches your briefing.