Sparse Readout Prism: Explaining Logit-Lens Scores in Features Instead of Tokens

A language model's prediction of its next token develops across layers, and lens methods track this process by decoding intermediate hidden states into tokens. But a lens reading reflects both the hidden state and the readout (the unembedding matrix) used to decode it. Many lenses are fit on a corpus, and we show that two lenses differing only in their fitting corpus can report different tokens for the same hidden states. We call this dependence corpus conditionality. To examine readout structure independently of the fitting corpus, we introduce Sparse Readout Prism (SRP), which decomposes the readout using only its weights and expresses any token logit or logit difference as a sum of contributions from sparse readout features. This reveals readout features as a new unit of analysis for lens readings, exposing structure that token identities can obscure and enabling comparisons across tokens, contexts, layers, and lenses. Replacing the original readout with SRP's sparse approximation reconstructs 8.9-17.3 percentage points more of the tested logit differences than the strongest of six baselines built on geometric relations among readout rows. Ablating features shifts logit differences in proportion to their SRP contributions. Although token readings vary with the fitting corpus, the dominant readout feature remains stable. Because SRP uses no corpus in its construction, it provides a control independent of the fitting corpus for lens analyses.

Beyond Scores: Understanding LLM-as-a-Judge Mechanisms in Summarization Evaluation

LLM-based evaluators of natural language generation (NLG) quality are widely deployed as scoring tools and as automated training signals, yet the internal procedure by which they assign a rating remains poorly understood. We investigate this procedure mechanistically through an eight-attack perturbation taxonomy across the Readability and Adequacy dimensions of NLG quality, a generation pipeline that produces paired clean and corrupt summaries with controlled error intensity and explicit token-level modification maps, and a four-experiment battery of causal tracing, logit-lens vocabulary projection, and attention-head knockout applied to Themis (Llama-3-8B) and Prometheus (Mistral-7B). Both evaluators implement a structured, coherent evaluation pipeline operating in two stages: below layer 15, attention performs local error comparison and routes the result to the final input position; above it, the MLP cascade integrates the signal and writes the rating, with the decision crystallizing in the residual stream at a sharp late layer (L = 26 on Themis, L = 25 on Prometheus). Furthermore, a base-model control at the same scale (Llama-3-8B) reproduces the routing architecture and crystallization but not the stage separation, isolating the two mechanisms that fine-tuning specifically installs, suppression of below-L15 MLP contribution at the last position and a two-layer advance of the crystallization depth, indicating that fine-tuning sculpts an existing substrate rather than building the pipeline from scratch. We release the source code and data at https://github.com/himil-v/judge-mech

VisLens: Single-Pass Interpretable Visual Search for Multimodal LLMs

Multimodal large language models (MLLMs) struggle with fine-grained Visual Search, the task of locating small or rare objects in high-resolution images. Existing remedies fall into two families: (1) Training-free methods based on attention or confidence scores are accurate but slow, since they require multiple MLLM queries per example. (2) Reinforcement Learning (RL) trained tool-use models are faster at inference but opaque, since their tool calls remain uncontrollable and hard to interpret. To overcome this, we propose \emph{VisLens} (Visual Focus via Logit Lens), a Visual Search method built on the logit lens, which decodes the semantics held in a hidden state by projecting it through the LLM head. VisLens further uses a lightweight tuned-lens that maps early hidden states into the final hidden state space, so visual tokens can be read out from early layers. These tokens are matched to target words in the query to generate a crop of the relevant region, which is fed back in alongside the original image to produce the final answer. The whole process, from decoding to the final answer, completes in a single forward pass without repeated queries. VisLens matches or exceeds prior baselines while delivering a substantial latency advantage, running $8.5$--$9.9\times$ faster than Thyme and up to $22.2\times$ faster than training-free multi-pass search methods.

Can We Read the Mind of an Audio LLM? A Verbalizable, Multilingual Middle-Layer Workspace

An audio language model is a black box in a specific way: we see what it says, never what it works out on the way there, and chain-of-thought monitoring helps only if the model writes its reasoning down. Reading a base Qwen3-Omni with a logit lens at the audio-token positions, we find that the answer to a spoken question becomes legible - in words - in the model's middle layers, before it emits any token. Five findings follow. (1) The readout carries concepts in neither the question, the options, nor the model's own transcription: on a clip whose verbatim transcription is empty garbling, it reconstructs Watergate and scandal, passes through the role president, and resolves to Nixon - a hidden multi-hop chain, read with no chain-of-thought. (2) The content is language-agnostic: one audio-inferred concept surfaces in several scripts at once, and 38% of top-1 readouts are Chinese on English inputs. (3) It is paralinguistic: given the same clip as audio and as the model's own emotion-free caption, the audio mind forms the sound source, speaker role, or affect that the caption discards, and answers correctly more often. (4) The audio-driven signal is absent at the input, turns on about a tenth of the way into the network, separates most cleanly from the text prior in the middle band (35-80% of depth), and activation patching shows it is causally used and committed before the last fifth of the layers. (5) Deleting single layers maps the pipeline: reading the sound in is localized to the entry layers and answer delivery to the output layer, while retrieval is distributed across the interior. Throughout, a waveform-swap control - identical text, only the sound changed - isolates the audio-driven signal from a prior over the printed options. This is a qualitative account of what an audio model works out before it speaks: the quantities are controls, not benchmark scores.

Reading Between the Dots: Decoding Hidden Computation across Filler Tokens

Frontier LLMs can perform multi-step reasoning over content-free filler tokens like dots or counting sequences, producing correct answers with no visible chain-of-thought (CoT). This is a limit case for behavioral oversight, where surface tokens carry no information about the underlying reasoning. But hidden from the output is not the same as hidden from us. On four task families (fact retrieval, parallel numeric composition, string manipulation, and in-context computation), two open-weights frontier models (DeepSeek V3, Kimi K2) compute over filler tokens in a structured, legible way: attention routes the question through the filler region to the answer, logit-lens readouts show retrieved facts emerging early and their composition crystallizing in late layers, and KV-cache transplants at filler positions causally swap outputs between examples. We introduce an unsupervised decoding pipeline that takes only hidden states as input and recovers intermediate values with 80-95% accuracy (best LLM judge) across both models and all four tasks, without ground-truth labels or training. Hidden computation that defeats behavioral CoT monitoring is, on these tasks, directly readable from the residual stream, suggesting monitorability is a property of the model's full computational trace, not just its surface tokens.

Interleaved Speech Language Models Latently Work In Text

Speech language models (SLMs) have been extensively studied, with the common paradigm incorporating text data and pre-trained text LMs. A leading approach is speech-text interleaving in which models are trained over sequences containing both speech and text tokens, aiming to boost even speech-only capabilities. Yet the way these two modalities interact in the model latent space remains unclear. In this work, we analyze interleaved speech-text LMs from different model families and sizes through the scope of the logit lens to provide such insight. We reveal that these models go through an implicit transcription phase in which the text token of the spoken word becomes decodable in intermediate layers, despite not being trained for speech recognition. The transcription of the word appears as one of the top candidate words for as much as 77\% of the data. Following this stage, the models proceed to predict the next word in the text space before transforming back to the speech domain. We finally analyze the role of interleaving data, and initializing from text LMs in eliciting this behavior, as well as seeing how this correlates with spoken knowledge abilities. Our analysis sheds light on the internal mechanisms underlying the relationship between speech and text modalities and could shape SLM optimization.