Alignment Collapse Under KV Cache Quantization: Diagnosis and Mitigation

• URL: http://arxiv.org/abs/2606.09864v1
• Date: Mon, 01 Jun 2026 02:02:20 GMT
• Title: Alignment Collapse Under KV Cache Quantization: Diagnosis and Mitigation
• Abstract summary: Key-value ( KV) cache quantization is widely used to reduce Large Language Model (LLM) inference memory.<n>In this study, we explore alignment preservation under KV cache quantization.<n>We find that low-bit quantization can silently destroy safety alignment.
• Score: 6.129872931808218
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Key-value (KV) cache quantization is widely used to reduce Large Language Model (LLM) inference memory, yet existing evaluations solely focus on measuring perplexity and accuracy without assessing the safety impact. In this study, we explore alignment preservation under KV cache quantization. Across eleven instruction-tuned models (3.8B-72B) and five benchmarks (1,894 prompts), we find that low-bit quantization can silently destroy safety alignment: Mistral-7B loses 15.2% of its refusals at only 1.03x perplexity, and no universal safe bit-width exists, with sharp model-specific phase transitions invisible to standard metrics. We identify that the root cause is geometric: safety features occupy a low-dimensional activation subspace 10^2-10^3x more vulnerable to quantization noise than the full representation space perplexity averages over. Inspired by this observation, we propose Per-Channel Reduction (PCR), a diagnostic that classifies each model into one of three mechanistic failure modes: outlier-crushes-safety, where safety lives in non-outlier channels collaterally damaged by outlier-driven scale factors; outlier-as-safety, where safety overlaps outlier channels and finer granularity cannot rescue it; and multi-layer dilution, where safety is distributed across many layers and per-layer fixes fail. PCR predicts the correct mitigation direction on all nine primary models and one held-out model from an independent family using 20 calibration prompts. PCR generalizes across unseen prompts, models, and production quantizers, including KIVI with up to 97.2% recovery, succeeding where attention-based allocation methods fail. The resulting training-free protocol, requiring approximately 35 GPU-minutes, recovers up to 97% of lost alignment at minimal memory overhead, addressing vulnerabilities confirmed in production vLLM serving with FP8 KV cache on NVIDIA GPUs.

Related papers

• The Structure of Quantization Damage in LLMs: Why the Next Bit Should Be Spent Globally [0.0]
  Post-training quantization (PTQ) is widely used to reduce the cost of serving large language models (LLMs)<n>We study where quantization damage occurs and how to allocate a small additional precision budget.
  arXiv  Detail & Related papers  (2026-09-01T17:53:41Z)
• SemKV: Semantic Mixed-Precision KV Cache Quantization Guided by the Quality Cliff for Long-Context LLM Inference [3.2938220347051175]
  We show that uniform KV quantization on a fractional-bit grid does not degrade gracefully.<n>We show that replacing the affine base with a distortion-optimized quantizer (TurboQuant-MSE) prunes the cliff in every protocol tested.
  arXiv  Detail & Related papers  (2026-08-28T22:19:47Z)
• From CVE to CWE: Syscall-Based HIDS Generalisation [41.99844472131922]
  In operational settings, defenders need to recognise new exploits of an already known type of weakness.<n>We empirically examine whether a one-class anomaly detector trained on the normal behaviour of a set of CVEs that share a Common Weaknession (CWE) class generalises to a different, unseen CVE inside the same class.
  arXiv  Detail & Related papers  (2026-06-21T16:34:39Z)
• Clipping Bottleneck: Stabilizing RLVR via Stochastic Recovery of Near-Boundary Signals [83.0127582612634]
  Near-boundary Rescue (NSR) is a minimal, plug-and-play modification that retains out-of-bound tokens to recover lost signals.<n>NSR substantially improves training stability and delivers consistent gains over strong baselines such as DAPO and GSPO.
  arXiv  Detail & Related papers  (2026-05-21T16:45:31Z)
• Runtime-Certified Bounded-Error Quantized Attention [0.0]
  We present a tiered KV cache architecture that enables runtime-certified attention.<n>A two-term error decomposition yields per-head, per-step bounds on (i) attention distribution distortion from key quantization and (ii) value reconstruction error.<n>Across PG-19, NIAH, and RULER benchmarks, the system matches dense FP16 KV quality within noise for language modelling and retrieval tasks.
  arXiv  Detail & Related papers  (2026-05-20T08:04:40Z)
• Measuring Five-Nines Reliability: Sample-Efficient LLM Evaluation in Saturated Benchmarks [45.86413490112477]
  Large language models (LLMs) are used in reliability-sensitive applications.<n>Estimateing rare failure probability with tight confidence bounds requires prohibitively large LLM inference sizes.<n>We propose to learn a sampling distribution concentrated on failure-prone inputs via the cross-entropy method.
  arXiv  Detail & Related papers  (2026-05-11T20:23:44Z)
• Continual Calibration: Coverage Can Collapse Before Accuracy in Lifelong LLM Fine-Tuning [6.908972852063454]
  Uncertainty reliability can degrade earlier and more sharply than top-1 performance.<n>We study this empirically by measuring conformal coverage and calibration error on sequentially fine-tuned models.<n>We propose calibration replay, a lightweight post-hoc procedure that maintains a task-specific buffer and refits a task-specific conformal threshold.
  arXiv  Detail & Related papers  (2026-04-27T03:03:38Z)
• The Illusion of Equivalence: Systematic FP16 Divergence in KV-Cached Autoregressive Inference [2.954681536906518]
  KV caching is a ubiquitous optimization in autoregressive transformer inference.<n>We show that FP16 KV cache inference is fundamentally non-equivalent to recomputation.
  arXiv  Detail & Related papers  (2026-04-16T15:59:40Z)
• Conformal Risk Control for Safety-Critical Wildfire Evacuation Mapping: A Comparative Study of Tabular, Spatial, and Graph-Based Models [0.0]
  We present the first application of conformal risk control (CRC) to wildfire spread prediction.<n>Standard thresholds capture only 7-72% of true fire spread, but CRC eliminates this failure uniformly.<n>Both models with CRC achieve approximately 95% fire coverage while flagging only approximately 15% of total pixels, making them 4.2x more efficient than LightGBM.
  arXiv  Detail & Related papers  (2026-03-20T21:05:13Z)
• BadCLIP++: Stealthy and Persistent Backdoors in Multimodal Contrastive Learning [73.46118996284888]
  Research on backdoor attacks against multimodal contrastive learning models faces two key challenges: stealthiness and persistence.<n>We propose BadCLIP++, a unified framework that tackles both challenges.<n>For stealthiness, we introduce a semantic-fusion QR micro-trigger that embeds imperceptible patterns near task-relevant regions.<n>For persistence, we stabilize trigger embeddings via radius shrinkage and centroid alignment.
  arXiv  Detail & Related papers  (2026-02-19T08:31:16Z)
• Lazy Layers to Make Fine-Tuned Diffusion Models More Traceable [70.77600345240867]
  A novel arbitrary-in-arbitrary-out (AIAO) strategy makes watermarks resilient to fine-tuning-based removal. Unlike the existing methods of designing a backdoor for the input/output space of diffusion models, in our method, we propose to embed the backdoor into the feature space of sampled subpaths. Our empirical studies on the MS-COCO, AFHQ, LSUN, CUB-200, and DreamBooth datasets confirm the robustness of AIAO.
  arXiv  Detail & Related papers  (2024-05-01T12:03:39Z)
• KVQuant: Towards 10 Million Context Length LLM Inference with KV Cache Quantization [67.74400574357472]
  LLMs are seeing growing use for applications which require large context windows, and with these large context windows KV cache activations surface as the dominant contributor to memory consumption during inference.<n> Quantization is a promising approach for compressing KV cache activations; however, existing solutions fail to represent activations accurately in sub-4-bit precision.<n>Our work, KVQuant, facilitates low precision KV cache quantization by incorporating several novel methods.
  arXiv  Detail & Related papers  (2024-01-31T18:58:14Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.