Pith. sign in

REVIEW 11 cited by

Relative representations enable zero-shot latent space communication

Not yet reviewed by Pith; the record is open.

Re-run · record.json Download PDF Read on arXiv ↗

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2209.15430 v2 pith:IVKMT2P6 submitted 2022-09-30 cs.LG cs.AI

Luca Moschella , Valentino Maiorca , Marco Fumero , Antonio Norelli , Francesco Locatello , Emanuele Rodolà This is my paper · ORCID
classification cs.LGcs.AI
keywords latentdataspacerepresentationstrainingarchitecturescommunicationneural
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Neural networks embed the geometric structure of a data manifold lying in a high-dimensional space into latent representations. Ideally, the distribution of the data points in the latent space should depend only on the task, the data, the loss, and other architecture-specific constraints. However, factors such as the random weights initialization, training hyperparameters, or other sources of randomness in the training phase may induce incoherent latent spaces that hinder any form of reuse. Nevertheless, we empirically observe that, under the same data and modeling choices, the angles between the encodings within distinct latent spaces do not change. In this work, we propose the latent similarity between each sample and a fixed set of anchors as an alternative data representation, demonstrating that it can enforce the desired invariances without any additional training. We show how neural architectures can leverage these relative representations to guarantee, in practice, invariance to latent isometries and rescalings, effectively enabling latent space communication: from zero-shot model stitching to latent space comparison between diverse settings. We extensively validate the generalization capability of our approach on different datasets, spanning various modalities (images, text, graphs), tasks (e.g., classification, reconstruction) and architectures (e.g., CNNs, GCNs, transformers).

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 11 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. You Had One Job: Per-Task Quantization Using LLMs' Hidden Representations

    cs.CL 2025-11 reject novelty 6.0 of 10

    TAQ estimates per-layer importance from hidden representations and output sensitivity on task calibration data to allocate mixed precision in a training-free PTQ setting, outperforming task-agnostic baselines on accur...

  2. Latent Space Alignment for AI-Native MIMO Semantic Communications

    cs.LG 2025-07 conditional novelty 6.0 of 10

    Joint MIMO precoder/decoder optimization for latent space alignment outperforms disjoint semantic alignment and channel equalization in simulations.

  3. Navigating the Latent Space Dynamics of Neural Models

    cs.LG 2025-05 conditional novelty 6.0 of 10

    Autoencoders implicitly define a latent vector field whose attractors encode the model's memorized and generalized knowledge, enabling data-free probing and out-of-distribution detection.

  4. Grounding Functional Similarity by Invariance-Aware Model Stitching

    cs.LG 2025-05 conditional novelty 6.0 of 10

    FuLA, a task-agnostic stitching objective that aligns intermediate features through the frozen end network, is claimed to be a more reliable functional similarity metric than task-based stitching.

  5. Calibrating Pre-trained Language Classifiers on LLM-generated Noisy Labels via Iterative Refinement

    cs.CL 2025-05 conditional novelty 6.0 of 10

    SiDyP improves classifiers trained on LLM-generated noisy labels by retrieving likely true labels from embedding-space neighbors and iteratively refining them with a simplex diffusion model, reporting average gains of...

  6. Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models

    cs.LG 2025-05 conditional novelty 6.0 of 10

    ITDA dictionaries, built greedily from poorly reconstructed activations and decomposed by matching pursuit, match some SAE reconstruction performance at 100x lower training cost and enable SOTA cross-model layer simil...

  7. How Far Do Simple Transformations Translate Across Text Embedding Models?

    cs.LG 2026-08 conditional novelty 5.0 of 10

    Simple linear translators between text embedding models work only for architecturally and training-similar pairs, so embedding spaces are not universally related by such maps.

  8. What are you sinking? A geometric approach on attention sink

    cs.LG 2025-08 reject novelty 5.0 of 10

    Attention sinks in transformers are reinterpreted as geometric reference frames, with three architecture-dependent types: centralized, distributed, and bidirectional.

  9. RIS-aided Latent Space Alignment for Semantic Channel Equalization

    cs.LG 2025-07 conditional novelty 5.0 of 10

    RIS-aided joint physical and semantic channel equalization, solved by alternating optimization or neural networks, outperforms separate alignment-and-transmission baselines in MIMO semantic communication simulations.

  10. Discovering Chunks in Neural Embeddings for Interpretability

    cs.LG 2025-02 conditional novelty 5.0 of 10

    Recurring 'chunks' in neural embeddings can be extracted, predict input patterns, and be perturbed to steer a model's outputs.

  11. Frame-Based Zero-Shot Semantic Channel Equalization for AI-Native Communications

    cs.NI 2025-07 conditional novelty 4.0 of 10

    Using Parseval-frame projections onto shared anchor features, a receiver can approximately reconstruct the latent vectors of an unseen, independently trained encoder; a Lyapunov scheduler then allocates bandwidth, CPU...

Pith tools