Stage-Replay Divergence Follows the KV Cache: Fixed-Prefix Precision Controls and Bidirectional Cache Transplantation

• URL: http://arxiv.org/abs/2607.28495v1
• Date: Thu, 30 Jul 2026 16:41:40 GMT
• Title: Stage-Replay Divergence Follows the KV Cache: Fixed-Prefix Precision Controls and Bidirectional Cache Transplantation
• Abstract summary: Stage-replay reconstructs intermediate token prefixes and treat fresh-prefill continuation as continuation from the decoder state that originally reached the prefix.<n>A matched 200-item experiment compares retained live cache with one-shot prefill of identical integer tokens and places an exact replica on both sides.
• Score: 51.56484100374058
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Stage-replay diagnostics reconstruct intermediate token prefixes and treat fresh-prefill continuation as continuation from the decoder state that originally reached the prefix. We audit that assumption at a whole reasoning-stage boundary in a Qwen2.5-derived system. A matched 200-item experiment compares retained live cache with one-shot prefill of identical integer tokens and places an exact replica on both sides. In BF16, replicas remain exact while the constructions differ on 166 suffixes and 20 correctness labels; the accuracy difference is only one point (paired 95% CI [-3.5, +5.5]). A fixed-prefix 2x2 holds all 200 token states constant while crossing construction and precision. The BF16 disagreements recur, whereas FP32 produces no decoded disagreement (95% Wilson upper bound 1.88%). A prospective bridge makes token-by-token incremental and retained live caches bit-exact on 12/12 rows; an all-200 saved-ledger audit reproduces every retained trajectory and comparison fingerprint. Bidirectional transplantation of all 48 key/value layers makes every tested divergent continuation follow its cache donor, both on a selected set at the primary checkpoint (24/24) and an outcome-blind replication at a later checkpoint (43/43). Exact-token replay can therefore be repeatable without preserving live-state fidelity. On the tested states, boundary K/V cache is a causally sufficient carrier of the divergent trajectory, while numerical precision moderates its behavioral expression.

Related papers

• Two tests of phase-structure features for transition prediction [0.0]
  This report examines whether phase-derived features improve endpoint prediction over a combined baseline in two settings.<n>Study 1 froze a contradiction-category pipeline before sealed scoring.<n>Study 2 developed fifteen treatments on blocks b0-b4 using 1,415 eligible answer-change comparisons and twenty-repeat, five-fold grouped cross-validation.
  arXiv  Detail & Related papers  (2026-08-31T20:29:18Z)
• CARD: Calibration via Agreement in Reverse Diffusion for Out-of-Domain MRI Segmentation [14.844162865265673]
  We propose CARD, which maps the temporal aggregate of disagreement to a temperature field applied per pixel across all classes, so that confidence changes while the segmentation does not.<n>Across cardiac, prostate and brain MRI shifts, CARD lowers calibration error in 45 of 49 comparisons against the strongest baseline in each setting.
  arXiv  Detail & Related papers  (2026-08-26T07:23:08Z)
• Same Agent, Different Answers: A Repeat-Aware Audit of Corpus-Induced Answer Churn in Retrieval-Augmented QA [0.4351760066527804]
  A retrieval-augmented QA system can return different answers after an index expansion even when its requested model identifier, prompt, retrieval policy, evidence depth, rendering, and exposed generation controls are held fixed.<n>We call the hidden phenomenon accuracy-blind answer churn and introduce the emphSnapshot Compatibility Audit, which estimates excess answer churn by subtracting same-snapshot repeat disagreement from cross-snapshot disagreement.
  arXiv  Detail & Related papers  (2026-08-24T06:40:09Z)
• A Pre-Specified Construction-Confirmation Test of Operation-Level Causal Transfer Across Finite Isomorphic Symbolic Domains [0.0]
  We ask if the hidden-state difference between two operations is estimated separately for each source input, does adding that difference move the model toward the corresponding recipient answer?<n>The design compares this input-specific intervention with wrong-operation, norm-matched random, and no-op controls, and separates candidate construction from an independently isolated confirmation split.
  arXiv  Detail & Related papers  (2026-08-16T15:37:26Z)
• From Expert Reduction to Behavioral Divergence: Tracing Numerical State through Sparse MoE Inference [0.0]
  Expert-reduction orders can produce observably different sparse-MoE executions.<n>We isolate this effect in native DeepSeek-V4-Flash by freezing local MoE state and varying only aggregation semantics.
  arXiv  Detail & Related papers  (2026-07-30T12:07:56Z)
• Looping Is Not Reliability: State-Bound Evidence and Typed Revision Contracts for Agentic Code Repair [36.56438114281786]
  We study the gap between finding a correct patch and retaining, verifying, and submitting it.<n>We derive an evidence-bound typed loop contract and instantiate its mechanically enforceable subset in a reference implementation.
  arXiv  Detail & Related papers  (2026-07-27T16:05:23Z)
• Which Values Do LLMs Confuse? A Schwartz-Based Recognition Study [32.505127447635864]
  We study controlled top-1 recognition over Schwartz's ten basic values.<n>We evaluate 21 instruction-tuned LLM runs under a fixed ranked-response protocol.<n>We find that models often locate the correct motivational region while ranking close alternatives unstably.
  arXiv  Detail & Related papers  (2026-07-22T15:22:14Z)
• CUSUM-Shaped Inference-Time Monitoring and Targeted Re-Decoding for Quantized Small Language Model Reasoning [0.6999740786886536]
  MGT-B maps overlapping windows of pre-sampling uncertainty and degeneration features to position-conditional empirical tail probabilities.<n>We audit whether the effect persists on problem identities first observed after the manual choice of log threshold h = 10.<n>Neither analysis is confirmatory, and the empirical factors are not established as a valid e-process or e-detector.
  arXiv  Detail & Related papers  (2026-07-22T13:34:17Z)
• Does Verbose Chain-of-Thought Really Help? In-Distribution Evidence that Content, Not Length, Matters [10.230437172157474]
  Chain-of-thought prompting improves LLM reasoning, but the source is contested.<n>We repeatedly sample each model on the same question and pair a shorter with a longer of its own natural generations that follow the same reasoning plan.<n>We find that the extra tokens leave accuracy essentially unchanged for every independently-trained reasoner.<n>We ask whether two traces expressing the same semantic content produce different outcomes when one is more verbose.
  arXiv  Detail & Related papers  (2026-06-29T11:06:46Z)
• Efficient Visual Pointing for Embodied AI:Agent-Driven Data Synthesis, Cross-Block Attention, and Iterative Correction [55.11480729304395]
  We describe our PointArena 2026 solution, which achieves 77.2% overall accuracy and ranks second on the benchmark.<n>The ap proach targets three failure modes. First, agent-driven syn thesis builds large semantic and anchor-relative candidate pools.<n>Second, a determinis tic steerable-data pipeline creates a verified 10,000-sample main set, plus reserve samples, using masks, templates, and path verification.
  arXiv  Detail & Related papers  (2026-06-29T06:39:03Z)
• Re-feeding Is Not Replaying: Measuring Replay Noise in Counterfactual Token-Credit Estimation [0.0]
  Per-token counterfactual credit estimation asks which token in a language-model rollout caused the final answer to be right or wrong.<n>We measure what that assumption costs on a stock inference engine, with a three-pass design.<n>Across six configurations and three models, re-feeding changes the credit estimate at rates 14-28 percentage points above the replica floor.
  arXiv  Detail & Related papers  (2026-06-14T06:09:16Z)
• Ghost: Plausible Yet Unlearnable Trajectories via On-Manifold Substitution for Next-POI Privacy [34.07056031726073]
  A publisher who releases check-in trajectories inadvertently publishes a strong predictor of every user's future locations.<n>We address this risk by generating unlearnable trajectories, perturbed sequences that yield victim models with degraded next-Point-of-Interest (next-POI) accuracy on clean test inputs.<n>We propose Ghost, a manifold-aligned framework whose perturbations look like plausible human check-in sequences yet leave no learnable signal behind.
  arXiv  Detail & Related papers  (2026-06-02T14:31:14Z)
• LaTER: Efficient Test-Time Reasoning via Latent Exploration and Explicit Verification [35.08680804423239]
  Chain-of-thought (CoT) reasoning improves large language models (LLMs) on difficult tasks, but it also makes inference expensive.<n>We propose Latent-Then-Explicit Reasoning (LaTER), a two-stage paradigm that first performs bounded exploration in a continuous latent space.<n>LaTER projects final-layer hidden states back to the input embedding space, preserves the latent KV cache, and uses entropy and model-native stop-token probes to decide when to switch.
  arXiv  Detail & Related papers  (2026-05-08T06:23:58Z)
• Adaptive Consensus in LLM Ensembles via Sequential Evidence Accumulation: Automatic Budget Identification and Calibrated Commit Signals [0.0]
  DASE produces a commit-type routing partition that generalises across benchmarks.<n> Adaptive stopping, not injection bandwidth, drives accuracy.<n>Injection-based methods exhibit an inverted-U accuracy-vs-inference trajectory.
  arXiv  Detail & Related papers  (2026-05-05T19:24:10Z)
• Capture-Quiet Decomposition: A Verification Theorem for Chess Endgame Tablebases [51.56484100374058]
  We present the Capture-Quiet Decomposition (CQD) for verifying Win-Draw-Loss labelings of chess endgame tablebases.<n>The theorem decomposes every legal position into exactly one of three categories -- terminal, capture, or quiet.<n>We validate CQD exhaustively on all 35 three- and four-piece endgames (42 million positions), all 110 five-piece endgames, and all 372 six-piece endgames -- 517 endgames in total.
  arXiv  Detail & Related papers  (2026-04-09T07:22:24Z)
• Fact or Facsimile? Evaluating the Factual Robustness of Modern Retrievers [34.31192184496381]
  Dense retrievers and rerankers are central to retrieval-augmented generation (RAG) pipelines.<n>We evaluate how much factual competence these components inherit or lose from large language models (LLMs) they are based on.<n>For every embedding model, cosine-similarity scores between queries and correct completions are significantly higher than those for incorrect ones.
  arXiv  Detail & Related papers  (2025-08-28T04:13:51Z)
• vCache: Verified Semantic Prompt Caching [95.16654660556975]
  This paper proposes vCache, the first verified semantic cache with user-defined error rate guarantees.<n>It employs an online learning algorithm to estimate an optimal threshold for each cached prompt, enabling reliable cache responses without additional training.<n>Our experiments show that vCache consistently meets the specified error bounds while outperforming state-of-the-art static-threshold and fine-tuned embedding baselines.
  arXiv  Detail & Related papers  (2025-02-06T04:16:20Z)
• Prune Spatio-temporal Tokens by Semantic-aware Temporal Accumulation [89.88214896713846]
  STA score considers two critical factors: temporal redundancy and semantic importance. We apply the STA module to off-the-shelf video Transformers and Videowins. Results: Kinetics-400 and Something-Something V2 achieve 30% overshelf reduction with a negligible 0.2% accuracy drop.
  arXiv  Detail & Related papers  (2023-08-08T19:38:15Z)

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.