# Mask-Guided KV Cache Eviction in Block Diffusion Language Models

arXiv: https://arxiv.org/abs/2610.06996 · HTML: https://mlsanity.com/paper/2610.06996/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Discrete and text diffusion, one of 14 directions in Diffusion models.
- Builds on 26 of its 29 references that are in the map.
- First posted on arXiv 4 Oct 2026.

## Placed in

- Diffusion models / [Discrete and text diffusion](https://mlsanity.com/diffusion/discrete-and-text-diffusion/)
- Efficient LLM inference and serving / [KV cache optimization](https://mlsanity.com/inference/kv-cache-optimization/)

## Builds on (most cited first, 26 of 26)

- [Efficient Streaming Language Models with Attention Sinks](https://mlsanity.com/paper/2309.17453/) (2023)
- [Large Language Diffusion Models](https://mlsanity.com/paper/2502.09992/) (2025)
- [Simple and Effective Masked Diffusion Language Models](https://mlsanity.com/paper/2406.07524/) (2024)
- [H2O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Models](https://mlsanity.com/paper/2306.14048/) (2023)
- [Discrete Diffusion Modeling by Estimating the Ratios of the Data Distribution](https://mlsanity.com/paper/2310.16834/) (2023)
- [KIVI: A Tuning-Free Asymmetric 2bit Quantization for KV Cache](https://mlsanity.com/paper/2402.02750/) (2024)
- [KVQuant: Towards 10 Million Context Length LLM Inference with KV Cache Quantization](https://mlsanity.com/paper/2401.18079/) (2024)
- [Fast-dLLM: Training-free Acceleration of Diffusion LLM by Enabling KV Cache and Parallel Decoding](https://mlsanity.com/paper/2505.22618/) (2025)
- [Block Diffusion: Interpolating Between Autoregressive and Diffusion Language Models](https://mlsanity.com/paper/2503.09573/) (2025)
- [Fast-dLLM v2: Efficient Block-Diffusion LLM](https://mlsanity.com/paper/2509.26328/) (2025)
- [No Token Left Behind: Reliable KV Cache Compression via Importance-Aware Mixed Precision Quantization](https://mlsanity.com/paper/2402.18096/) (2024)
- [Expected Attention: KV Cache Compression by Estimating Attention from Future Queries Distribution](https://mlsanity.com/paper/2510.00636/) (2025)
- [Sparse-dLLM: Accelerating Diffusion LLMs with Dynamic Cache Eviction](https://mlsanity.com/paper/2508.02558/) (2025)
- [ParallelBench: Understanding the Trade-offs of Parallel Decoding in Diffusion LLMs](https://mlsanity.com/paper/2510.04767/) (2025)
- [SparseD: Sparse Attention for Diffusion Language Models](https://mlsanity.com/paper/2509.24014/) (2025)
- [Mask Tokens as Prophet: Fine-Grained Cache Eviction for Efficient dLLM Inference](https://mlsanity.com/paper/2510.09309/) (2025)
- [Lookahead Q-Cache: Achieving More Consistent KV Cache Eviction via Pseudo Query](https://mlsanity.com/paper/2505.20334/) (2025)
- [More Tokens, Lower Precision: Towards the Optimal Token-Precision Trade-off in KV Cache Compression](https://mlsanity.com/paper/2412.12706/) (2024)
- [LookaheadKV: Fast and Accurate KV Cache Eviction by Glimpsing into the Future without Generation](https://mlsanity.com/paper/2603.10899/) (2026)
- [DreamReasoner-8B: Block-Size Curriculum Learning for Diffusion Reasoning Models](https://mlsanity.com/paper/2606.19257/) (2026)
- [Where Matters More Than What: Decoding-aligned KV Cache Compression via Position-aware Pseudo Queries](https://mlsanity.com/paper/2603.11564/) (2026)
- [RDKV: Rate-Distortion Bit Allocation for Joint Eviction and Quantization of the KV Cache](https://mlsanity.com/paper/2605.08317/) (2026)
- [NPU Design for Diffusion Language Model Inference](https://mlsanity.com/paper/2601.20706/) (2026)
- [FlashBlock: Attention Caching for Efficient Long-Context Block Diffusion](https://mlsanity.com/paper/2602.05305/) (2026)
- [MAGE: All-[MASK] Block Already Knows Where to Look in Block Diffusion LLM](https://mlsanity.com/paper/2602.14209/) (2026)
- [GroupKV: Hierarchical KV Cache Management for Long-Context Diffusion LLM Inference](https://mlsanity.com/paper/2609.17573/) (2026)

## Abstract (arXiv)

Block diffusion language models keep a large key-value (KV) cache throughout generation and attend to it at every denoising step, limiting both memory capacity and generation speed. Reducing these costs requires deciding which past tokens to use for denoising the current block (selection) and which to keep in memory for future blocks (eviction). We propose MaskAhead, a training-free method that solves both tasks with a single mask-query-based ranking mechanism. Current-block masks guide selection, while probes of upcoming masked blocks guide eviction. Both rank KV entries by their estimated contribution to the attention output. Our quantized variant, Q-MaskAhead, computes selection and attention directly from low-bit KV, largely preserving the selected entries. Experiments on Fast-dLLM-v2, DreamReasoner, and LLaDA2.0-mini cover long-generation reasoning, long-prompt question answering, and needle-in-a-haystack retrieval. On long-prompt QA, MaskAhead reduces KV memory by 9.5× on average with a 1.2-point mean F1 loss relative to dense inference. Q-MaskAhead increases the reduction to 20.1× with a 2.3-point mean F1 loss. In a batch-32 systems profile, MaskAhead achieves 1.23× end-to-end and 1.68× decode-stage speedups over dense inference.