Lost in State Space: Probing Frozen Mamba Representations

• URL: http://arxiv.org/abs/2605.00253v1
• Date: Thu, 30 Apr 2026 21:35:08 GMT
• Title: Lost in State Space: Probing Frozen Mamba Representations
• Abstract summary: Mamba's recurrent state h_t is a compressed summary of every token seen so far.<n>If we extract token-level outputs y_t at fixed patch boundaries, we obtain semantic sentence summaries for free.<n>We compare four strategies for extracting frozen sentence representations from a pretrained Mamba-130M backbone.
• Score: 0.5156484100374058
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Mamba's recurrent state h_t is, by construction, a compressed summary of every token seen so far. This raises a tempting hypothesis: if we extract token-level outputs y_t at fixed patch boundaries, we obtain semantic sentence summaries for free, with no pooling head, no fine-tuning, and no [CLS] token. We test this hypothesis carefully. Across five benchmarks (SST-2, CoLA, MRPC, STS-B, IMDb), we compare four strategies for extracting frozen sentence representations from a pretrained Mamba-130M backbone under a strict frozen-feature probing protocol, using three random seeds where computationally feasible. The results do not support the hypothesis: patch boundary readouts do not consistently outperform simple mean pooling. We identify and quantify two structural pathologies: severe anisotropy (mean pairwise cosine similarity 0.9999, std 0.000044) and representational collapse in the raw final SSM state (MCC = 0.000 on CoLA across all three seeds, confirmed via confusion matrix). We further propose orthogonal injection, a modified recurrence that constrains new information per

Related papers

• Accurate Ensembles, Fragile Narratives: Multi-Scale Stacking and a Fidelity Audit of LLM-Generated Explanations for Credit Risk [0.0]
  Credit scoring increasingly relies on models whose decision logic cannot be read off their parameters.<n>A common proposal closes that gap with a language model: compute feature attributions, hand them to an LLM, and let it write the rationale.<n>We build such a system end to end and test whether the second half of the promise holds.
  arXiv  Detail & Related papers  (2026-08-08T13:22:14Z)
• Mean-to-Score Discrete Diffusion: Posterior-Mean Denoisers for Score Entropy [17.48959208263029]
  We introduce emphmean-to-score (M2S), which predicts a clean-token posterior mean and converts it to the score through an exact kernel-dependent linear map.<n>For uniform corruption, it maps the probability simplex onto the bridge polytope; for absorbing-mask corruption, the resulting objective recovers MD4 exactly.
  arXiv  Detail & Related papers  (2026-07-23T14:37:30Z)
• Significance-First Splitting: Aligning Treatment Heterogeneity Detection with Honest Estimation [0.0]
  Estimating heterogeneous treatment effects (CATE) requires simultaneously detecting effect modification and quantifying estimation uncertainty.<n>Existing tree-based methods make an uneasy trade-off: significance-based approaches identify subgroup interactions directly but lack valid inference.<n>We introduce a hybrid algorithm that fuses significance-based splitting with honest sample-splitting and cross-validation.
  arXiv  Detail & Related papers  (2026-07-04T19:50:35Z)
• Bayesian Wind Tunnels for Model Selection [0.4779196219827507]
  We show that transformers can perform exact Bayesian filtering within a fixed hypothesis class.<n>We then identify a sharp perceptual access condition: when the discnative statistic requires arithmetic, model selection succeeds with integer tokens.<n>A stationarity control confirms the operative factor: stable semantics, not integer identity, enable circuit compilation.
  arXiv  Detail & Related papers  (2026-07-01T04:05:08Z)
• CaricHarmony: Contrastive Diffusion Paths for Identity-Preserving Caricature Synthesis [49.596677723190886]
  Sketch-based caricature synthesis suffers from a fundamental failure mode.<n>Identity and shape conditions are combined in diffusion models, causing collapse toward bland portraits or unrecognizable distortions.<n>We present CaricHarmony, the first training-free method that explicitly resolves this contamination through parallel uncontaminated diffusion paths.
  arXiv  Detail & Related papers  (2026-06-11T22:57:59Z)
• Convergence Theory for Iterative LLM-Based Neural Architecture Search: A Parametric Cross-Entropy Framework with Closed-Form Proxy Reliability [48.83701310501069]
  Large language models (LLMs) are increasingly used as generators in iterative neural architecture search (NAS)<n>We model iterative LLM-NAS as a parametric Cross-Entropy (CE) method over executable programs.<n>We prove six results: (1) iterative LLM fine-tuning on elite architectures is equivalent to the CE update restricted to the LLM family; (2) expected architecture quality is monotonically non-decreasing across cycles; and (3) elite-set probability converges to a fixed point at a geometric rate.
  arXiv  Detail & Related papers  (2026-05-28T15:45:19Z)
• Sparse Regression under Correlation and Weak Signals: A Reproducible Benchmark of Classical and Bayesian Methods [1.6679662639178268]
  We benchmark six sparse regression methods on synthetic data.<n>Bayesian methods win on prediction error (MSE 72 vs. 108-267), and the Horseshoe delivers near-nominal 95% coverage (94.8%)<n>For variable selection, Lasso and Spike-and-Slab tie at F1 0.47, making Lasso the practical default when posteriors are not needed.
  arXiv  Detail & Related papers  (2026-04-04T15:46:44Z)
• Token Coherence: Adapting MESI Cache Protocols to Minimize Synchronization Overhead in Multi-Agent LLM Systems [0.0]
  Multi-agent LLM orchestration incurs synchronization costs scaling as O(n x S x |D|) in agents, steps, and artifact size under naive broadcast.<n>I argue this pathology is a structural residue of full-state rebroadcast, not an inherent property of multi-agent coordination.<n>I construct the Artifact Coherence System (ACS) and prove the Token Coherence Theorem.
  arXiv  Detail & Related papers  (2026-03-16T12:20:06Z)
• On the Probability of First Success in Differential Evolution: Hazard Identities and Tail Bounds [0.0]
  We study first-hitting times in Differential Evolution (DE) through a conditional hazard frame work.<n>For the L-SHADE algorithm with current-to-$p$best/1 mutation, we construct a checkable algorithmic witness event $mathcal L_t$ under which the conditional hazard admits an explicit lower bound.
  arXiv  Detail & Related papers  (2026-01-16T18:24:24Z)
• Your Absorbing Discrete Diffusion Secretly Models the Bayesian Posterior [0.0]
  We prove that the expected denoiser output under the forward corruption distribution recovers the true posterior.<n>We introduce an inference-time ensemble that runs K independent denoising passes and aggregates both posterior means and variances.
  arXiv  Detail & Related papers  (2025-07-10T09:42:47Z)
• Near-Optimal Clustering in Mixture of Markov Chains [74.3828414695655]
  We study the problem of clustering $T$ trajectories of length $H$, each generated by one of $K$ unknown ergodic Markov chains over a finite state space of size $S$.<n>We derive an instance-dependent, high-probability lower bound on the clustering error rate, governed by the weighted KL divergence between the transition kernels of the chains.<n>We then present a novel two-stage clustering algorithm.
  arXiv  Detail & Related papers  (2025-06-02T05:10:40Z)
• Online Covariance Estimation in Nonsmooth Stochastic Approximation [14.818683408659764]
  We consider applying approximation (SA) methods to solve nonsmooth variational inclusion problems.<n>Our convergence construction establish the best-known for statistical estimation methods.
  arXiv  Detail & Related papers  (2025-02-07T20:16:51Z)
• 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)
• Optimal Clustering with Bandit Feedback [57.672609011609886]
  This paper considers the problem of online clustering with bandit feedback. It includes a novel stopping rule for sequential testing that circumvents the need to solve any NP-hard weighted clustering problem as its subroutines. We show through extensive simulations on synthetic and real-world datasets that BOC's performance matches the lower boundally, and significantly outperforms a non-adaptive baseline algorithm.
  arXiv  Detail & Related papers  (2022-02-09T06:05:05Z)
• Provably Breaking the Quadratic Error Compounding Barrier in Imitation Learning, Optimally [58.463668865380946]
  We study the statistical limits of Imitation Learning in episodic Markov Decision Processes (MDPs) with a state space $mathcalS$. We establish an upper bound $O(|mathcalS|H3/2/N)$ for the suboptimality using the Mimic-MD algorithm in Rajaraman et al ( 2020) We show the minimax suboptimality grows as $Omega( H3/2/N)$ when $mathcalS|geq 3$ while the unknown-transition setting suffers from a larger sharp rate
  arXiv  Detail & Related papers  (2021-02-25T15:50:19Z)

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.