Reading

Grounded language-model pipelines can be divided into three stages: selecting an object, retrieving passages for it, and using that evidence to answer. If the selected object must reach the reader, losing it breaks the handoff. Benchmark recall checks the dataset-linked object, which can differ. We audit 600 HybridQA questions across three selector families. On 1,463 resolvable records where the selected object matches the dataset-traced passage, exact key lookup and exact title matching return the object every time. With every ranked rule given the same decoded selected title, body-only BM25 omits it on 389 records (26.6%) at cutoff five, while hybrid retrieval with reranking omits it on 14 (1.0%). The two identities differ on 329 of 1,792 resolvable records. With original-question rankings, their top-five checks disagree on 106 records (5.9%). Frozen reader comparisons associate the aligned object's presence with 28.6 to 31.0 points higher exact match. In a deliberately selected 64-item cohort, removing that passage sharply lowers exact match, while removing a similar-length comparison passage does not reproduce the drop. We release the Returned-Object Profile (ROP), an executable record of the target, returned-ID field, cutoff, membership rule, and complete expected population, with data and an offline replay.

Compressed context is usually carried as human-readable text or as rendered images that must be decoded, even when its consumer is a language model. We introduce LatentPress, which writes conversational histories and long documents into a third representation: continuous memory tokens that a frozen decoder reads directly through its input-embedding interface, with no text reconstruction at inference. A small reader-matched writer compresses $4$-$16\times$ while training only an adapter (4.2M-26.2M parameters, $\sim\!0.1\%$ of the decoder). On LongMemEval, LatentPress reaches $0.504$ accuracy at $7.70\times$ compression versus $0.490$ for uncompressed evidence, outperforming text summaries (0.184) and OCR-based compression (0.426 to 0.312). On LongBench-QA, in-domain writers match or exceed raw-context reading at $4$-$8\times$ compression, while $16\times$ trails raw. Writing takes 43ms per conversation, roughly an order of magnitude faster than text summarization or OCR reconstruction, and reading is $5$-$9\times$ faster than raw context or cached OCR. We validate the interface under two transfer settings, zero-shot from UltraChat to LongMemEval memory QA and from LongMemEval-derived QA to unseen LongBench document domains, establishing direct soft tokens as a practical machine-facing context interface beyond text and vision. The implementation of the experiments could be found at: https://github.com/HJSang/LatentPress .

Large language models are increasingly capable in general, but their utility can remain modest in niche or understudied areas. One approach to address this limitation is to specialise existing models through additional training on target-domain corpora. In this work, we investigate such continued pre-training for adapting large language models to Swedish journalism, using a high-quality dataset that we curate from millions of news articles. To evaluate the adaptation efficacy, we also construct a novel domain-specific benchmark that covers six editorial tasks. Through full and parameter-efficient fine-tuning across two model sizes, we find that continued pre-training yields benefits in the target domain, but only when paired with experience replay to mitigate forgetting. We observe consistent enhancements in the models' generation quality and factual knowledge, but not their proficiency in discriminative tasks. Exploring a training-free method to facilitate instruction following, we see further improvements, but exclusively for models trained with low-rank adaptation. Crucially, we demonstrate the importance of targeted evaluation in the adaptation process, as an existing Swedish benchmark largely fails to capture the models' in-domain performance gains.

Standard language proficiency tests rely on linguistic tasks such as vocabulary, grammar and reading comprehension quizzes. An alternative, cognitively motivated approach, introduced in Berzak et al. (2018), proposed instead to predict language proficiency from behavioral traces of eye movements in reading. In this work, we validate and extend this approach from single sentences to more naturalistic reading of contextualized passages in English as a second language, new proficiency measures, prediction models, and reading in an information seeking regime. We find that the approach is effective in all these evaluations. We further address two key open questions on eye movement based proficiency testing: (1) potential scoring biases that reflect the proximity of the reader's native language to English, which may undermine validity, and (2) its reliability. We find that eye movement based proficiency scores are indeed biased towards L1s that are linguistically closer to English. We propose a score debiasing method which effectively remedies this issue. The reliability analyses suggest that eye movement proficiency scores are more reliable than standard language proficiency scores. Overall, our results strengthen and broaden the empirical foundations for future eye movement based language assessment technologies.

Purpose: Keyphrases are statistically and semantically important textual units that can also attract readers' attention during comprehension. However, existing keyphrase extraction (KPE) studies mainly focus on improving textual representation while largely overlooking human reading behavior. This study examines whether lightweight webcam-based eye-tracking features can improve KPE from Chinese academic abstracts in Library and Information Science (LIS). Methodology: To address the limited availability of eye-tracking data for Chinese academic reading, we developed a lightweight webcam-based data collection platform using the open-source SearchGazer library and constructed the Chinese LIS Eye-Tracking Corpus (CLIS-ET). Three character-level eye-tracking features, first fixation duration (FFD), fixation number (FN), and total fixation duration (TFD), were incorporated into KPE models to evaluate their effects on extraction performance. Findings: Eye-tracking features consistently improved KPE performance. The combination of FN and TFD achieved the best results on the Att-BiLSTM+CRF model, indicating that readers' fixation behavior provides useful signals for identifying keyphrases in academic abstracts. Originality/value: This study introduces a cost-effective webcam-based eye-tracking approach for KPE and presents CLIS-ET, a Chinese academic eye-tracking corpus containing FFD, FN, and TFD features. The results demonstrate the value of incorporating human reading behavior into keyphrase extraction. Dataset and code: https://github.com/yan-xinyi/ET_AKE and https://github.com/yan-xinyi/Reading_ET_System.

Multi-step language-model agents repeatedly process growing interaction histories, leading to substantial context costs. Vision--text compression reduces these costs by rendering history as images, but the resulting modality shift creates a marked capability gap. Through controlled evaluations of history recovery, matched-state decisions, and complete trajectories, we show that this gap cannot be explained by OCR quality alone. Visual-history agents exhibit systematic drift in action selection, query formulation, stopping, and evidence use, revealing an agentic policy gap. We introduce \textbf{CAPS}, a two-stage \textbf{C}ross-modal \textbf{A}gentic \textbf{P}olicy \textbf{S}elf-distillation framework that uses the same model's stronger text-history policy to supervise its visual-history counterpart. Offline trajectory self-distillation transfers successful text-policy behavior to visual-history inputs, while online policy self-distillation provides dense supervision on states visited by the visual-history policy during reinforcement learning. On SearchQA, CAPS improves over AgentOCR by 5.0% and 3.4% with 3B and 7B backbones, respectively. On full-history ALFWorld, the corresponding gains are 15.6% and 14.5%. Across settings, CAPS reduces average memory-context cost by up to 63.3% and peak cost by up to 83.4% relative to matched text-history policies. These results show that explicit cross-modal policy self-distillation can preserve agent capability under vision--text compression. Our code will be made publicly available in a future release.

Existing measures of how much a text is about a concept read the surface of the text: dictionary word shares, topic proportions, embedding similarities. They score the words a text uses, not the judgment a reader forms about it. Recent work has shown that a gap exists in what Large Language Models (LLMs) know internally versus what they express in their response. This paper asks whether that internal knowledge, read by monitoring the activations of frozen, out-of-the-box LLMs, can stand in for task-specific fine-tuning when measuring concept content, and which extraction method reads it best. We extract such measures via the Recursive Feature Machine (RFM) algorithm and via linear probing, and compare these against an embedding baseline, surface baselines, and the same model's own answer to the question. We demonstrate the approach on financial text, a domain studied extensively and served by established annotated resources, using a human-annotated Environmental, Social and Governance (ESG) dataset. The best linear probe comes within 0.6 percentage points of a fine-tuned domain classifier's accuracy without any task-specific fine-tuning, and outscores the same model's own answer to the question in eleven of twelve comparisons, so the activations carry concept content the response does not report. The simple probe consistently beats the RFM concept vectors, which in turn provide what classification alone does not: a continuous score intended to reflect how strongly a concept is present in a text, whose validation awaits graded labels.

We present Stockmark-Nemotron-3-Nano-Omni-JapanDocReader, a Japanese document understanding model built from Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16. The central goal of this work is structured document parsing via capability injection and forgetting control: we inject Japanese structured document parsing capability into a reasoning-oriented multimodal model while preserving its document VQA capability as much as possible. We study parsing-centric SFT, which uses only structured document parsing data; mixed SFT, which combines structured document parsing and VQA data; and parsing-centric RL, which optimizes structured parsing with a task-level reward. Our experiments show that parsing-centric SFT substantially improves structured document parsing performance but causes measurable VQA forgetting. Mixed SFT mitigates this forgetting while preserving nearly the same structured parsing performance. Applying DAPO-based parsing-centric RL on top of the mixed SFT checkpoint further improves structured document parsing beyond the SFT ceiling, producing the final released model. The training data is constructed with a data engine consisting of two complementary synthetic streams: a Japanese Document VQA Stream and a programmatic structured document parsing stream. We also discuss reward design and variance-based prompt filtering for continuous structured document parsing rewards, highlighting their importance for making RL effective in long-reasoning structured document parsing tasks.

Social media videos often communicate meanings that go beyond their visible actions, captions, or speech. A mundane clip may become humorous, ironic, or satire only through the interaction of multimodal cues and cultural context, making such content a difficult test case for video-language models. In this paper, we introduce \textit{DrivelHub+}, a benchmark for evaluating whether models can infer the implicit, non-linear, and rhetorically layered meanings of social media videos that appear nonsensical on the surface but convey deliberate pragmatic meanings. DrivelHub+ consists of 1,000 videos collected from social media, each annotated with a human-written implicit narrative explanation. Unlike conventional video understanding tasks focused on recognition or description, we present a benchmark that targets contextual multimodal reasoning. We evaluate current video-language models from two perspectives: explanation, where models must explain the pragmatic comprehension of a video in natural language; and representation, where we adapt reasoning-as-retrieval to test whether model representations align videos with their corresponding implicit narratives in both video-to-text and text-to-video retrieval. Our benchmark provides a diagnostic setting for measuring the gap between multimodal perception and pragmatic comprehension, asking whether current models can move beyond describing what is shown to inferring what is meant.

High-resolution pixels and crop or zoom tools give multimodal large language models the ability to inspect an image, but they do not provide a reliable task-conditioned policy for deciding where to inspect. Q-CueGraph makes this decision explicit. It maps a question and an image representation to budgeted, coordinate-level observations for a frozen reader. Text-rich images use a reusable OCR/layout graph; natural-image search instantiates query-conditioned visual nodes behind the same selection, composition, and budgeting interface. Optional utility refinement learns which candidate crops the frozen reader can use from training-answer correctness, without region-box supervision. With a frozen Qwen2.5-VL-7B reader, Q-CueGraph reaches 0.833 accuracy on V*Bench versus 0.696 for full-image inference from a 19% image-area budget, and reaches 92% of full-image ANLS on InfographicVQA from about half the image area. Across six benchmarks, explicit observation is most valuable when evidence is localizable, the question discriminates its location, and resolution limits full-image reading.

Search and database engines still store text as UTF-8, a format built for humans. But the systems that increasingly read and write that text (embedders, rerankers, and language-model agents) work with token IDs, not characters, so every access pays to translate between the two. As agents become the primary readers and writers of stored text, we argue for token-native storage: keep the text as the model's own byte-pair-encoding (BPE) token IDs. Packing r50k IDs as uint16 already beats UTF-8 by 2.25x on English with no compression, and an entropy coder on top reaches 3.30x. Across six tokenizers and three corpora (English, code, Hindi), compressing token IDs matches or beats every byte codec, even a corpus-trained zstd dictionary. Two findings sharpen the case. BPE numbers tokens by merge order instead of frequency, and re-ranking by frequency lets a plain integer codec (streamvbyte) recover most of the entropy coder's ratio while decoding ~7x faster, a near-free change to how AI labs publish vocabularies. And because a model reads token IDs, not text, a token-native store hands over the IDs directly instead of re-tokenizing on every read. The only requirement is that reader and writer share a tokenizer, and different model families often use different ones today, so we argue for standardization: a published, shared vocabulary, the way ASCII and UTF-8 standardized text.

Large language models often show users a final response and a short reasoning summary while the full reasoning trace stays hidden. We introduce an observability ladder that holds each completed run fixed and varies only what a reader inspects to judge whether the answer is correct: the response, a self-summary the model writes from the trace, the trace itself, and internal signals, each with and without the prompt. Across three benchmarks and five open-weight Qwen3 and gpt-oss models, we train matched linear correctness predictors on each access level. Without the prompt, summaries carry most of the trace's ranking signal (mean AUROC 0.774 versus 0.813) and add +0.156 over the response alone. With the prompt visible, the summary's gain collapses to +0.019, while the trace still adds +0.041. Even at equal length, the trace's last words predict correctness as well as summaries, or slightly better, and carry denser and more discriminative uncertainty and self-correction cues. On MMLU-Pro questions with both correct and incorrect runs, linear summary readers are near chance and trace readers retain only modest signal, both with and without the prompt (prompt-withheld AUROC 0.503-0.545 versus 0.544-0.590). With the prompt withheld, a GPT-5-mini reader recovers substantially more signal from both summaries and traces on gpt-oss-20b, and even then the trace keeps a small +0.034 advantage. Much of the linear readers' trace signal is associated with length. In the common case where users already hold the prompt, summaries are less helpful than the full trace for monitoring correctness. Monitorability is thus a joint property of the display and the reader, so any monitorability claim, including for faithfulness, should specify both.

A benchmark score means nothing without knowing what a trivial method achieves and what the best possible method could achieve. We construct both bounds for a task with a rare kind of ground truth: predicting which sentences a crowd of readers -- highlighting for their own purposes, unpaid, uninstructed, and blind to each other -- marked in 120 web documents. The floor is naive truncation (lead); the ceiling is a split-half oracle: half the crowd predicting the other half. The gap between them is +0.2028 AP [+0.1698, +0.2342, domain-clustered], and three findings structure it. First, the gap is semantic: position and length features recover 5% of it. Second, frontier language models reach 35-53% of it zero-shot -- far above classical baselines, far below the crowd; a state-of-the-art prompt compressor (LLMLingua-2) lands below the floor, indistinguishable from random selection. Third, an unweighted cross-vendor fusion of five frontier rankings plus a position prior reaches 60%, beating the best single model by +0.0159 [+0.0044, +0.0269; Holm p=0.019] -- a gain that survives ablation of its best member, split-half arm selection, prompt paraphrase, and label, gate, and seed perturbations, and was CONFIRMED by a pre-registered replication on 217 independent documents (+0.0179, Holm p=0.042). Finally, the bracket compresses: distilling the fusion into one open-weight 8B student that reads the whole document retains 90% of the fusion's edge and reaches statistical parity with the strongest single frontier model (+0.0070 [-0.0068, +0.0200]), where a local-context student retains only 63% -- the crowd's signal lives in document-level structure, and the cheapest known improvement is to ask several different models and average.

Claims that language models homogenise are usually measured against human judgements collected for the study, which makes the human side an artifact of the design: a crowdworker given the model's instruction is running the model's prompt. We measure convergence against a human reference nobody built for the purpose -- 2,523 reader mark sets across 120 web documents, produced by people highlighting for their own reasons on a platform where the overlay of others' marks is off by default. Agreement is the overlap between two size-matched sentence sets minus the overlap expected when each is resampled within its own depth-and-length bands. The null's calibration is demonstrated, not asserted: every pair involving a random baseline lands within 0.006 of zero. On the median document each party names 14 sentences of 70; two readers share 4.1 and two models 8.7. Across 18 model arms spanning 11 vendors, 3 countries and both weight regimes, the median of 153 model pairs is +0.093 against a human yardstick of +0.040, and 99 sit entirely above the human interval. Two frontier models from rival labs reach +0.203, twice what GPT-4o agrees with itself on a second call. The effect is not determinism, prompt wording, procedure, vendor or routing, and it is graded: the smallest models agree at the human level. No model agrees with readers detectably more than a reader does, and at equal depth and length no surface feature separates their choices. The multiples are procedure-dependent and the ordering is not: models are cut to their sharpest set while a reader's is a random draw from what they marked, and blunting the models alike halves the gap without closing it. Tested out of sample on four models released after this analysis, against predictions fixed beforehand, none clears the human interval. A population simulated from several models is not several populations.

Context compression discards most of a document before a language model reads it, and is normally evaluated by downstream task accuracy - which makes another model the judge of what mattered. Naturalistic social highlighting offers a non-circular reference: many people independently marking passages on the same page. But the obvious metric, the fraction of crowd-marked sentences a compressor keeps, is confounded twice: crowd marks are front-loaded and crowd-marked sentences are longer, so any method favouring early or long sentences scores well regardless of readers. We remove both by matching each marked sentence against unmarked sentences of the same document at equal relative depth and equal within-document length rank, and we calibrate every estimator on synthetic nulls built from position and length alone - a step that matters, since depth-only stratification returns a false positive on 20-36% of nulls containing no effect. On 120 web documents (at least 12 independent readers each), a language-model importance ranking keeps 38.4% of crowd-marked sentences against 19.9% of their matched neighbours: an enrichment of +0.196 [+0.148, +0.239], at p = 0.0005 under an exact randomization test that assumes nothing about clustering, and replicated cross-vendor. Naive truncation, whose keep rule is position, correctly falls to +0.003. To give the number a scale: scored identically, on the same budget, against a crowd label recomputed to exclude them, a single human reader reaches +0.182 - indistinguishable from GPT-5.4 (+0.002 [-0.081, +0.088]) and below Claude Opus 5. Classical methods are not null - Luhn's 1958 heuristic reaches +0.088 - so reader selection is partly recoverable by counting words; conditioning additionally on lexical centrality removes only 0.010, so the agreement is not centrality. We also report that a claim in our own prior work does not reproduce on this corpus.

When it comes to the proper classification of ancient coins with respect to their time and issuer, the textual inscriptions on these coins, also known as legends, are of paramount importance. These legends consist of alphabets or characters still used in English. This paper addresses image based character recognition on ancient Roman Republican coins via a deep learning based object detection strategy. However, legends on these coins pose high variation due to non-uniform placement, primitive inscription techniques, and wear and tear. Additional challenges include inconsistent imaging conditions such as illumination, orientation, and scale. To accommodate these, we gathered a novel large-scale dataset of 5,654 Roman Republican coin images, manually annotated with 21 character labels, totaling 38,808 annotations. For recognition, we use You Only Look Once (YOLO) variants: YOLOv3, v4, v5, v7, and v8. YOLOv7-Large achieves the best mAP50 of 90.4%, followed by YOLOv7-Extended and YOLOv7-xl with 90.2% and 90.1%, respectively.

Activation Oracles (AOs) are language models trained to answer natural-language questions about another model's internal activations. They offer a flexible interface for reading hidden information from model states, especially when relevant information is internally represented but absent or incomplete in visible behavior. However, AOs are themselves learned systems: their answers are shaped by training data, objectives, and learned reporting behavior, rather than being neutral readouts of represented information. We study this in a controlled Taboo Word Guessing setting, where subject models are fine-tuned to internally use a hidden concept while avoiding direct disclosure. Contrary to the expectation that an AO trained on such a subject becomes a specialist reader, we find that fine-tuned AOs can become concept-specific anti-readers: they selectively fail to recover the concept persistently present during their own training. This failure is not simply explained by absence of the concept from the subject or oracle representations: the target remains decodable inside the oracle, while LogitLens and layer-ablation analyses indicate that the failure arises in the AO readout pathway. Our results show that behavioral leakage, representation-level decodability, and AO-verbalizability can come apart, raising a reliability concern for learned interpretability interfaces.

Natural-language autoencoders score explanations of hidden activations by reconstruction: an explanation is deemed faithful if the activation can be regenerated from it. The test is structurally insensitive to individual false claims: if flipping a claim does not change the reconstruction, the claim is never penalized. We show the test is passed in two ways, neither faithful. On a released Qwen-2.5-7B verbalizer, explanations reconstruct well above chance while ~2% of specific claims are reconstruction-dependent, so the score tracks gist, not specific facts. Under exact synthetic ground truth, the standard recipe develops co-adapted private codes (false wording the reconstruction depends on) in 5/5 runs, and fixes that leave the target model unchanged do not help. We contribute two audit protocols, the grounded-vs-true cross and the evaluator swap, and RECAP (Readable Encodings via Co-trained Auxiliary Predictors): linear heads trained alongside the target model to keep designated content decodable. On RECAP-trained sandbox models, fresh verbalizers state the designated content truly and the codes vanish, at a +0.001-nat cost. This replicates on a pretrained Pythia-160M: the content becomes reliably probe-decodable, though a fresh verbalizer conveys it only in part (truth 0.44-0.46 vs a near-zero control). For interpretability, high reconstruction does not certify individual claims. For AI safety, RECAP makes designated internal content independently checkable against probes rather than asserted by prose a model can game: an independent probe scores the verbalizer's true claims above its false ones (AUC 0.96, vs 0.82 without RECAP). Against an adversary that edits an explanation to maximize the reconstruction score while lying (suppressing ~87% of its lie penalty), the RECAP probe still flags the lies (AUC 0.95) while the control probe collapses to chance (0.51).

Large language models can answer scientific questions, yet a correct output does not reveal whether the model represents or uses the governing physics. Here, using three open-weight Gemma 4 models (google/gemma-4-E4B-it, google/gemma-4-12B-it, google/gemma-4-31B-it) we identify three experimentally separable signatures of materials-science mechanism information: selective concept readability, relational encoding of qualitative constitutive orientation, and causal, context-dependent control of constrained engineering answers. We combine matched direct and Jacobian vocabulary readouts, option-free state geometry, a 60-law counterfactual benchmark and causal interventions. In 50 held-out materials descriptions, three independently fitted Jacobian lenses reproduced concept ranks, and target-free word sets from both readouts enabled blinded identification of 9 of 10 mechanism families. A separate 72-prompt benchmark produced mechanism-specific hidden-state neighborhoods, but an exact graph audit showed that this apparent physical organization was equally explained by numerical comparison. We therefore compared otherwise identical prompts in which only the direction of the physical input was reversed, asking whether the resulting hidden-state movement followed the supplied constitutive law. These state transformations ordered direct, physically neutral and inverse laws across 60 frozen relations and correctly oriented 39 of 40 directional laws, whereas lexical controls were near chance. Bidirectional interventions shifted answer probabilities toward or away from the physically appropriate outcome across all 12 matched cases, while counterfactual state patches transferred opposing decision signals across mechanisms and answer formats. Physical relationships were therefore more visible in controlled state changes than in absolute states alone.

Independent agents that reason in latent space can share computed state as key-value cache fragments rather than text. Merged by a conflict-free replicated data type, these fragments form a store that converges under any delivery order or duplication. Yet a later query, unknown at encode time, cannot reliably read the merged cache: colocated fragments interfere, so colocation is not addressability. MaSRead addresses the read to content. It routes through opaque keyed tag sets derived from fragment words and decodes each selected fragment under a hard attention mask that hides the rest. Under lexical connectivity, a graph walk reaches the fragments required by a multi-hop query. Across chain, pipeline, symmetric, hub, and natural-language stores, MaSRead recovers visited fragments in isolation, remains effective as unrelated fragments accumulate, and transfers to another model family. After routing, materialized decoding depends on fragment length rather than total store size; end-to-end work still includes store-dependent routing and one read per visited fragment. The limits are explicit: lexical routing can miss disconnected evidence, and answer composition remains bounded by the frozen reader. Thus a replicated latent store becomes selectively readable for later queries when the needed fragments connect to the query through content.

In multi-hop RAG evaluation, a top-k answer score can hide two different failures: the retrieval window may drop part of the support chain, or it may contain support in a form the adapted reader does not use well. We call this reader-facing form of retrieved evidence an evidence interface. Using three support-annotated multi-hop QA benchmarks, we compare matched adapted readers trained with raw context, retrieval windows, and gold-support diagnostic renderings. These comparisons distinguish support-availability failures from remaining reader-interface effects. Top-k windows become interpretable only after checking whether the complete annotated support chain survives: when it does, short ranked windows can match or improve over raw context; when it does not, missing support explains much of the loss. Gold support-first improves matched readers; on 2Wiki and MuSiQue, a support-supervised ranker raises coverage and recovers raw-context quality at lower prompt cost, while retaining gold headroom. Support-removal checks further show that the gains rely on exposed evidence, not only answer priors. On support-annotated evaluations, top-k answer scores should therefore be reported together with complete-support coverage.

Agentic retrieval-augmented generation (RAG) extends static RAG by allowing language models to iteratively reason, generate search queries, retrieve evidence, and predict answers. However, it remains challenging for models to decide when to retrieve, whether to use lexical matching or semantic similarity, and how to control context granularity to prevent irrelevant tokens from interfering with agent reasoning. In this paper, we introduce GRASP, a reinforcement learning (RL) framework for training agents to adaptively coordinate complementary retrieval tools during multi-step reasoning. GRASP provides the agent with semantic search, keyword search, and paragraph-reading actions, enabling it to retrieve sentence-level evidence and expand further context only when needed. We train the policy with a reward that jointly accounts for answer accuracy, grounded reading, complementary search, and turn efficiency. Experiments on multi-hop reasoning benchmarks show that GRASP improves both retrieval recall and downstream question answering performance compared with single-step retrieval, prompting-based agentic RAG, and RL-based retrieval baselines. Qualitative and ablation analyses show that the learned policy develops interpretable skimming and scanning behavior: it uses semantic search for broad exploration, paragraph reading for local verification, and keyword search for entity-specific evidence. These results suggest that learning to coordinate retrieval signals and context granularity is critical for agent's correct reasoning.

A transformer can be built from operators that are legible by construction -- bounded, named units that read as fuzzy set operations rather than dense activations -- but legibility must be pressed for during training, and the pressure has a failure mode. A crispness penalty meant to sharpen a bounded operator into a decisive detector instead collapses it into a dead constant. An identity, E[v(1-v)] = mu(1-mu) - var, shows why -- the penalty is a variance-minimizer blind to the difference between a live detector and a constant -- and names the fix: a per-channel variance floor, the target legibility metric written as a loss, which recovers both legibility and quality. A learned per-unit fraction then retires the hand-set reserved-GELU partition of prior work: given the choice the model keeps no unit as pure GELU and routes 87% of its load-bearing computation through crisp operators. The result is the most legible transformer we have built -- 78% of its feed-forward operands and 50% of its attention value channels are crisp-and-contextual detectors, and per-head legibility rises from 18% in shallow layers to 78% in deep ones. Read in the correct rotated per-layer frame, these units separate a clean detection (what a unit responds to) from a harder naming (what its output decodes to); and because the objective makes each unit crisp and sparse, edits to them are far more local -- 50-184x in the deep layers where the edit sites concentrate -- and can target explicit conjunctions a single neuron cannot express. Finally, a between-unit decorrelation pressure exposes a legibility dial: it trades a circuit's reuse for independence at no quality cost, turning concepts into single, surgically editable units and a prediction into a short explanation read off a handful of named operations. Quality holds at parity with a conventional baseline throughout.

Event-based lip reading has recently emerged as a promising direction for visual speech recognition, benefiting from the high temporal resolution and motion sensitivity of event cameras. However, existing methods typically perform spatial compression before sufficient temporal modeling, which may suppress sparse and localized motion trajectories that are crucial for distinguishing similar lip movements. Moreover, most current approaches optimize temporal representations mainly at the word-classification level, leaving the underlying articulatory structure weakly constrained. To address these limitations, we propose a temporally enhanced framework for event-based lip reading. First, we introduce Trajectory-Aware Differential Aggregation (TDA), which performs local temporal modeling at each spatial location before adaptive spatial aggregation. Second, we propose Viseme-Guided Aggregation (VGA), a unified temporal module composed of a CTC decoder and a viseme-guided gated aggregation branch, which injects viseme-aware sequence supervision and improves final temporal aggregation for word recognition. Third, we incorporate an EMA teacher--student training strategy to enhance robustness under strong event perturbations. Experiments on the DVS-Lip benchmark verify the effectiveness of the proposed design, and extensive ablation studies further validate the contributions of TDA, VGA, and teacher--student consistency. Qualitative decoding results also demonstrate that the proposed CTC-based temporal modeling learns meaningful viseme-aware structure from event streams.

Revised following peer review. We expanded baseline comparisons, corrected evaluation leakage and read-boundary handling, clarified the confidence-weighted loss, and added sensitivity analyses for pooling and region selection. We also expanded TCS failure-mode and limitations analyses, added a discussion section, and provided code and data links for reproducibility.

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.

Reading order inference remains a critical bottleneck in the digitization of complex historical manuscripts, where pages contain multiple spatially interleaved reading streams, the canonical example being the Glossa Ordinaria layout, in which a central text is surrounded by commentaries that wrap around it in non-rectangular, non-convex regions. We present a training-free, graph-based framework: each OCR text line becomes a node in a directed candidate-transition graph, edges are scored by a weighted additive ensemble of two lightweight language-model signals (causal language model conditional likelihood and BERT next-sentence prediction, NSP; a third sentence-embedding signal was evaluated but did not improve reading order), and the global reading order is recovered as a degree-constrained directed path cover. To avoid the cascading "edge-theft" failures of greedy edge selection, we propose a max-regret inference rule that prioritizes commitments with high opportunity cost. We evaluate on synthetic Glossa Ordinaria grid layouts, on 23 ALTO page geometries (10 historical source pages plus mirrored and flipped variants), and on a 140-page multi-column English subset of OmniDocBench, comparing our method against the canonical recursive XY-cut (PaddleOCR PP-StructureV3) and two LayoutReader variants (layout-only and text+layout) on identical inputs. On wrap-around Glossa layouts our method recovers 95% of ground-truth successor edges on average vs. XY-cut's 50%; on the OmniDocBench multi-column subset it reaches 88% macro edge accuracy versus XY-cut's 75% and LayoutReader's 25%. The LayoutReader baselines transfer poorly due to a word-level vs. line-level granularity mismatch. We additionally verify mirror-invariance under horizontal and vertical page reflections: Our method changes by less than 1 percentage point, classical XY-cut by 2 points, and LayoutReader-T by up to 8 points.

Rhetorical strategies and their influence on audiences are often studied through social media posts and comments. However, this focus overlooks the universal audience, which is the majority of readers who remain silent and do not explicitly express how a message affects them. This study investigates how two classical modes of persuasion, ethos and pathos, resonate in the silent audience's interpretations of meaning. Using a dataset of social media sentences paired with human-written interpretations, we label both sources for ethos and pathos and assess whether these rhetorical appeals are preserved. Our analyses show that interpretations diverge from the original sentences in 30% of cases, with rhetorically charged content eliciting greater variability than neutral content. We further find that ethos and pathos in original sentences can predict audience attitudes toward the author, underscoring the subtle ways rhetoric shapes perception beyond visible engagement.

Automatic speech recognition for children's reading remains underdeveloped for most African languages, including Bambara, despite its potential value for reproducible literacy assessment. We present an open-source system for assessing children's reading in Bambara, developed through an end-to-end process linking field data collection, benchmark construction, model adaptation, a reading application, and classroom validation. A mobile collection and assessment app was used to collect 55 hours of raw reading speech from 60 children, from which we construct a public benchmark for Bambara child-reading assessment. Fine-tuning experiments compare Soloni, a Bambara-adapted Fast-Conformer ASR framework with TDT and CTC decoders, with QuartzNet, a compact convolutional ASR architecture. The best Soloni model reduces WER from 0.42 to 0.22 and CER from 0.15 to 0.08, substantially outperforming QuartzNet on the isolated benchmark. The experiments further show that repeated readings of the same texts provide architecture-dependent benefits: they substantially improve QuartzNet but add only marginal gains for Soloni, while SpecAugment regulates training without exceeding the best unaugmented configuration. Disaggregated analysis identifies children under 10 as the main source of residual errors, motivating targeted collection from younger readers. Ten classroom trials supported continued use of the application.

The carbonized papyri from Herculaneum preserve the only large-scale library to survive from classical antiquity, but many unopened rolls remain unread because physical opening risks irreversible damage. X-ray computed microtomography ($μ$CT) and virtual unwrapping offer a non-invasive route to their texts, yet previous work on sealed Herculaneum scrolls has recovered only localized readings or limited surface regions. Here, using high-resolution phase-contrast $μ$CT acquired on the BM18 beamline at the European Synchrotron Radiation Facility (ESRF), together with improved computational unrolling and machine learning, we achieve the complete virtual unwrapping and reading of PHerc. 1667 under explicit coverage and papyrological-review criteria. This makes PHerc. 1667 the first Herculaneum papyrus to be fully digitally unrolled and read for extended scholarly study without physical opening. In PHerc. Paris 4, the optimized scan protocol makes ink directly visible in the tomographic volume, allowing three-dimensional ink segmentation and independent validation of surface-conditioned ink recovery. In PHerc. 139, we recover title and author-attribution evidence identifying the scroll as Philodemus, On Gods, Book 8. These results move virtual unwrapping of the Herculaneum scrolls beyond isolated demonstrations towards a scalable framework for systematic recovery of the still-unopened library.

AI translation of literary works is increasingly common. While the content may be rendered adequately, we do not know enough about how readers experience it in terms of immersiveness and literary effect-aspects poorly captured by automatic metrics or human evaluation targeting fluency and adequacy. We ask 15 avid readers to compare recently published human translations (HT) to machine translations (MT) generated with an agentic language model-based pipeline, for 15 recent novels in French, Polish, and Japanese translated into English. Readers evaluated approximately 8K-word excerpts in two conditions: immersive reading of the whole excerpt (30 comparisons) and close reading of 386 aligned HT-MT chunk pairs (772 comparisons), with two readers per book. Overall, readers find MT "fine", but prefer HT (descriptively at the excerpt level, 19/30, and significantly at the chunk level, 522/772) for its ease, clarity, and immersive nature. Readers' highlights show that MT's quality varies more within one book than HT's does. Crucially, readers cannot reliably tell the two apart (17/30 guess correctly) and tend to prefer the version they believe to be human. Automatic metrics, including LLM-as-a-judge approaches, fail to recover reader preferences and favor MT. We release LAIT (Literary AI Translation), a reader-centered evaluation dataset with 1K reader comments, 2K judgments and preference ratings, and 7.2K span-level annotations, along with our evaluation protocol and supporting interface.

Retrieval-Augmented Generation (RAG) compression papers often evaluate a compressor on one to three readers and treat the compressed evidence layer as evaluation-neutral. We show this assumption is false: fixed compression can raise average accuracy while hiding reader upgrades and reversing model rankings. Across 20 readers and ten domain-method settings over four QA benchmarks and one summarization benchmark, compression gain decreases with reader baseline (nine of ten settings significant, p < 0.05). Generic summarization flips 31% of pairwise model rankings on LongMemEval-S, and a fixed HotpotQA compressor hides 80% of the raw upgrade from Qwen 7B to GPT-4.1-mini. Two opposing forces explain this paradox: compression rescues weak readers by removing noise they cannot filter, and harms strong readers by dropping details they would have used. The pattern appears across structured compilation, generic summarization, three trained compressor families, query-focused summarization, and an external audit of nine published compression papers. We release ragscale, a toolkit built on 177,000 row-level compression transitions, so any compression paper can audit reader scaling with three readers in one day.

Most companies read their customer support data at scale using sentiment analysis, which measures how customers sound rather than whether they were satisfied with the result. We tested a richer alternative on 70,450 support conversations from a leading online fundraising platform: alongside tone, we used GPT-5.4 to estimate each customer's satisfaction and to flag whether they reported a concrete problem, then validated all three readings against the 1-to-5 ratings customers left on the conversations they rated. The satisfaction estimate tracked those ratings far better than sentiment did, correlating at 0.47 against 0.36 and flagging unhappy customers with far fewer false alarms. The structured read also sees what sentiment cannot: tone and satisfaction disagree in 44% of conversations, a single "Neutral" label hides everything from quietly satisfied customers to ones who quietly gave up, and the largest group of all is "tolerated friction," customers who are satisfied but still reporting a fixable problem, a standing issue that no sentiment-based dashboard can surface. The broader finding is that LLM-based annotation can capture far more than the tonality of a customer's language, offering strong potential for new business metrics grounded instead in the customer's state (whether they were satisfied) and the cause of their problem extracted directly from the raw textual data of interactions and feedback.

This work presents the design, implementation, and evaluation of a system for generating personalized reading content using Large Language Models (LLMs) combined with Retrieval-Augmented Generation (RAG). The proposed architecture consists of four modules: Input, RAG, Generation, and Judging and enables users to specify both a question and a target reading content complexity. RAG is employed to retrieve relevant information from the Internet, enriching and grounding the content produced by three modern LLMs: Meta LLaMA 4 Scout, LLaMA 3.1 8B Instant, and Google Gemma2 9B. Reading materials are generated using three prompting strategies (Chain-of-Thought, zero-shot, and few-shot), and the LLM-as-a-Judge module automatically evaluates answer quality and alignment with the desired readability level. Experimental results show that RAG consistently improves system performance across all models and prompting techniques, increasing relevance and particularly groundedness by up to 26-35 percentage points. Overall, the findings demonstrate that the RAG-augmented architecture effectively produces reading content tailored to user queries and desired textual complexity.

Prior work on a social web highlighter located individuality in selection -- which documents a person chooses to highlight -- but measured it cross-sectionally. We ask the temporal question: is a reader's selection signature a trait or a state? We freeze each reader's first six months of highlighting as a profile and track its own-vs-other advantage on their later selections at growing gaps (to 24+ months), with negatives drawn from the same calendar era -- so supply drift cannot masquerade as personal drift -- at a coarse global level and at a fine level whose negatives and controls come from the reader's own interest neighborhood; the anchor cell reproduces the prior cross-sectional level (+0.188 vs +0.169), validating the harness. Four results. Within the same users, the fine-layer advantage shows no statistically detectable paired decline at any horizon (6-12 month retention R = 1.00 [0.85, 1.18], n = 212; the farthest bin is compatible with a modest decline; the only contrast whose interval excludes zero is the coarse layer at 12-24 months, about 13%). The signal is not reducible to repeated domains (~90% survives excluding all profile sources). Within-person drift is slow (a recent-half profile beats the old half by +0.042). Prospectively, personal profiles -- even one built from a reader's earliest documents, median 20 months before evaluation -- rank their next reads at roughly 3x the AP of every simple non-personal prior tested. We use "trait" operationally (a stable signature under continued engagement); the scope is heavy, long-tenured readers of one platform, and exposure is not separable from choice.

A social highlighter's most useful signal -- which passages a crowd of readers marks -- exists only for documents people have already read. Can the aggregate crowd salience of a document be predicted from its text before its marks accumulate? Prior work on this data found that zero-shot language models recover highlight locations worse than a trivial lead (position) baseline, so we ask whether a model trained on the highlight corpus can beat that baseline. Using a pre-registered ladder of models and a by-document cluster bootstrap, we find a small but robust edge: a logistic ranker over sentence embeddings and positional/contextual features beats the lead baseline by +0.044 average precision (95% CI [+0.029, +0.058]; clears a pre-registered margin delta=0.03 in 97% of resamples, and stable across pipeline re-runs). Two unsupervised extractive baselines (centroid, LexRank-style centrality) lose to lead, and the trained model beats them by +0.108, so the edge is not recovered by generic unsupervised proxies -- it reflects learning from real reader marks. In product terms, precision@3 rises from 0.25 to 0.39 (+55% relative) and the model beats lead on 69% of documents. An ablation attributes the edge to the raw embedding (+0.014) and training augmentation (+0.010), each with a positive CI. The edge is not a temporal-generalization failure, and we find no evidence that content drift or near-duplicate leakage explains it. A standardized regression shows the advantage is governed mainly by document popularity (lower popularity, larger edge) and by label reliability. It nearly vanishes only on the most popular content; there it is the lead baseline that strengthens, not the model that weakens. Because our evaluation conditions on documents that eventually accumulated readers, these results are a retrospective cold-start simulation.

When many people highlight the same document, is the crowd a single consensus, or is it internally structured into reader sub-groups that mark different things -- and is that structure a stable property of a reader or of the document? Building on prior work showing an individual's within-document highlighting signal is a whisper while individuality lives in selection, we ask the group-level question on a co-readership platform using a margin-preserving curveball null. Experiment 1: within a document, readers form strong sub-groups -- pairs agree far beyond what shared salience, mark density, and sentence popularity predict (nearest-neighbour agreement z=+6.3, significant in 88% of documents). Under an eight-block region-preserving null, shared engagement with the same coarse regions of the document accounts for about 40% of this excess; the majority survives as finer reader-specific agreement (z=+3.6, 77% significant). So the within-document crowd is, in a descriptive sense, factional. Experiment 2: is that grouping a stable reader trait? Here we are honest about power. The cross-document split-half reproducibility of a pair's agreement is near zero pooled (+0.078 and 0.000 in two separately drawn samples), and a power calibration shows the test is informative only for pairs that co-read many documents. In the only informative high-overlap subset (k>=4), point estimates are positive but small-sample, imprecise across the separately drawn samples, never significant, and attenuate under the region-preserving null. We therefore leave cross-document stability unresolved: the data is consistent with anything from situational grouping to a weak-to-moderate stable reader trait. The crowd is factional within a document; whether its factions follow the reader across documents is, honestly, beyond our reach.

As newsrooms integrate generative AI, journalists face a disclosure challenge: how to communicate AI involvement in ways that maintain reader trust. Current practice offers two approaches: brief one-line labels or detailed disclosures specifying human oversight, editorial accountability, and error reporting mechanisms. Neither achieves journalists' goal of building trust through transparency. An existing controlled experiment with 34 news readers show that detailed disclosures trigger a \textit{transparency dilemma}, reducing trust rather than increasing it, and risk introducing dark patterns that readers scroll past with the illusion of transparency. One-line disclosures avoid this effect but can create an information gap, prompting readers to expend cognitive effort searching for signs of AI involvement that the disclosure indicates but does not explain. Yet readers are not rejecting transparency, they proposed disclosure designs centered on user agency: detail-on-demand interactions, proportional AI-ratio visualizations, outlet-level signals, and explicit "no AI" labels. I argue that this disconnect between what practitioners believe is responsible disclosure and what users actually need is a design problem for the HCI community.

Does personalizing what a reader sees pay off, and where does it stop? Using a social web highlighter and a co-readership identity control (the same document highlighted by many users, which holds document and topic fixed and asks whether a person's own history predicts their marks better than another reader's does), we map the shape and limits of personalization across reading altitudes. At the document altitude we give the clean, leakage-free, identity-controlled measurement that prior next-document evaluations could only upper-bound: a person's history identifies which documents in a co-reading neighborhood are theirs, with an own-versus-other gap of +0.169 against community negatives and +0.119 against topic-matched hard negatives (both highly significant); a content-based arm suggests the signal is not purely title-driven but is largely thematic. This is comparable to the span-level selection signal (+0.14) from our prior work: the selection signal is of comparable magnitude across altitudes (+0.12 to +0.17), most of it stable topic preference. At the sentence altitude, a two-stage personalized auto-highlight (an impersonal model proposes candidates, a personal model re-ranks them) does not improve on its impersonal baseline: two off-the-shelf zero-shot LLMs, including a frontier model, predict highlight locations worse than a lead baseline, and personal re-ranking is beaten by the salience order even on the highest-recall candidate pool, so the null is not merely a Stage-1 ceiling artifact. Measurable personalization appears primarily at the selection layer: modest (~+0.13), topic-dominated, with no reliable gain at the salience layer. We also surface a control-in-negatives bias that inflated our document gap to a spurious +0.227 until audited. Going beyond the shared salience layer may be better approached by aggregating individuals than by personalizing them harder.

In abdominal CT imaging, developing a low-dose, no-reference image quality assessment (No-reference IQA) model that mimics doctors' reading habits for evaluating CT image quality has significant practical value. This paper proposes a novel deep learning-based framework, ClinReadNet, whose design aligns with the clinical reading logic of radiologists: first, it introduces the Sobel ordinal quality network (SOQN) module, which can simultaneously focus on edge details highly relevant to image quality and the quality distribution pattern of the entire image, accurately matching the clinical image-reading judgment habit of "considering both local details and overall context"; second, the framework integrates the (shifted) window multi-scale temperature multi-head self-attention ((S)W-MTMSA) module, which further replicates the radiologists' image-reading process of shifting from overall scanning to local focusing, and accurately locks in regions of interest through multi-sharpness attention; third, it designs the hierarchical ranked probability score (HRPS) loss function, which combines the dual logics of coarse classification and fine classification, while paying attention to the distance information between grading labels, effectively improving the performance of image quality assessment. Experiments conducted on the LDCTIQAG2023 dataset show that the proposed method achieves the current state-of-the-art (SOTA) performance: the values of Pearson's linear correlation coefficient (PLCC), Spearman's rank-order correlation coefficient (SROCC), and Kendall's rank-order correlation coefficient (KROCC) reach 0.9507, 0.9554, and 0.8629 respectively, with the sum of their absolute values (Score) being 2.7690, outperforming existing methods.

Speech restoration through silent speech interfaces (SSIs) has emerged as a promising assistive technology for individuals with impaired or absent laryngeal voice production. Among non-invasive SSI modalities, surface electromyography (sEMG) and video-based lipreading provide complementary articulatory information, yet their integration for continuous speech synthesis remains underexplored. Moreover, existing multimodal approaches rarely address robustness to modality degradation or temporary sensor failure, limiting their applicability in realistic scenarios. In this work, we propose a masked multimodal speech synthesis framework that jointly leverages sEMG and lipreading signals through modality masking during training. Under multispeaker settings, the proposed approach reduces word error rate by up to 14 absolute percentage points compared to the strongest unimodal baseline. Experimental results not only show that masking strategies are critical for these performance gains and robustness under low-bitrate conditions, but also that they generalize better than degradation-specific data augmentations in the presence of modality absence conditions. Phone-level analyses further reveal complementary contributions across modalities, with particularly strong benefits for vowels and for specific consonant groups. Overall, these findings demonstrate the effectiveness and robustness of masked multimodal integration for silent speech synthesis, although adaptation to laryngectomized speakers remains an open research challenge.

ReadingMachine is a computational methodology for structured corpus reading that uses large language models to perform bounded reading operations over entire document collections. Rather than relying on retrieval or recursive summarization, the approach decomposes analysis into inspectable stages including insight extraction, semantic clustering, theme generation, and iterative omission detection. By delaying irreversible compression and explicitly tracking intermediate representations, the method prioritizes coverage, traceability, and preservation of disagreement across large corpora. The system is demonstrated on a heterogeneous corpus of 152 industrial policy documents, producing more than 17,500 extracted insights and a structured thematic map. ReadingMachine is released as an open-source experimental framework for large-scale qualitative synthesis and corpus analysis.

Visual speech recognition (VSR) models now surpass human lipreaders on benchmarks, but do such gains establish human-like visual speech perception? To explore this, we compare three VSR systems with human baselines on the MaFI word-level lipreading dataset using word, character, phoneme, and viseme-level metrics. Although models achieve higher overall accuracy, they succeed and fail on different words than humans. A text-only n-gram baseline given only a few initial phonemes rivals human lipreading. VSR word-level errors are consistently better explained by training word frequency than by the visual informativeness of words. Viseme accuracies, confusion matrices and human-model correlations further show that models gain most on visemes humans find hardest, and show much weaker dependence on visual clarity. Our work demonstrates that VSR systems rely primarily on language cues from training data rather than visual perception, failing to bind visual features into meaningful words.

Semantic association between a word and its context has been identified as an important component of reading comprehension, even when word predictability is accounted for. Recent research has highlighted the potential of language model ( LM) embeddings to quantify semantic association. Yet, embedding-based semantic association have been operationalized in a myriad of ways. In this study, we use embeddings from LMs to estimate semantic association on a corpus of joint electroencephalography (EEG) and self-paced reading of natural, Dutch texts. Semantic association is calculated in ten different implementations that vary the embedding model and context lengths. The effects of semantic association across the different implementations on the N400 and self-paced reading times are examined using Bayesian hierarchical models and Bayes factor. The results show that the choice of embedding model can alter the estimated effect of semantic association on both the N400 and self-paced reading times. Furthermore, the results demonstrate a promising potential of sentence embeddings for capturing semantic association, as only implementations relying on sentence embeddings indicate reliable results of semantic association beyond word predictability on both neural and behavioral measures. Together, these findings highlight the importance of methodological choices in quantifying semantic association.

Retrieval-augmented QA pipelines often route retrieved passages through an LLM \emph{rewriter} before a smaller reader, lifting F1 by tens of points on multi-hop benchmarks; this gain is typically credited to improved evidence quality. We ask whether that lift is causally driven by the gold answer string appearing in the rewritten context rather than by curation per se, using a controlled intervention audit. For each rewritten context we re-run the reader after one of four controlled edits to the compile output: removing the gold answer span, replacing a length-matched random non-answer span (placebo), or injecting the gold into rewrites where it was absent (at the prefix or at a midpoint sentence boundary). Across twelve completed (cell, baseline) intervention runs spanning three reader families (Qwen2.5-7B, Qwen3.5-35B, GLM-4.7), two datasets (HotpotQA, 2WikiMultihopQA), and three compiler arrangements (MA-only, MB-only, MA$+$verify), removing the gold answer drops reader F1 by $28$ to $64$ points beyond the length-matched placebo on paired \texttt{answer-in-compile} strata, and prepending the gold into rewrites that lacked it raises F1 by $+0.7$ to $+9.7$ points in $10$ of $12$ (cell, baseline) combinations. A companion five-sentinel audit shows the conventional single-\texttt{[MASK]} probe is itself sentinel-fragile: on 2Wiki it reports a $+4.12$~F1 ``non-leakage residual'' that flips to $-3.33$ to $-7.81$~F1 under four alternative sentinels and fails an equivalence test for three of those four ($1/4$~pass). We do not propose a new rewriter or mitigation; we release the intervention runner and the sentinel panel so that other rewriter-gain claims can be tested against the same standard.

Context-augmented language model systems often wrap supplied content with labels such as Reference:, Evidence:, Instruction:, Note:, or Example:, but the effect of these labels on reader-model behavior remains underexplored. We introduce a paired fixed-content probe over 500 MMLU-Pro items: each item receives the same misleading answer-bearing assertion under different discourse-role labels, and adoption is measured by whether the model outputs the injected wrong option. Across GPT-5.5, DeepSeek V4 Pro, Llama-3-8B-Instruct, and Qwen2.5-7B-Instruct, Misleading Adoption Rate shifts by 56-84 percentage points. Binding or source-like labels such as Instruction: and Reference: produce high adoption, whereas Example: consistently suppresses it. Paired tests, bootstrap intervals, final-instruction ablations, and Qwen final-step log-probability probes support a label-conditioned candidate preference. Boundary probes show where the effect weakens or persists: arithmetic tasks reduce adoption, passage-shaped external context preserves smaller label gaps, short-answer evaluation rules out option-letter copying, and nested-label conflicts suggest that illustrative framing can delimit adoption scope. A 200-case single-author manual audit confirms that the short-answer contrasts are stable under conservative adjudication. The resulting claim is bounded but practical: context-utilization and reader-side RAG benchmarks should report and control wrapper labels, because presentation choices can change measured reliance on supplied context.

Knowledge graphs over corpora of inter-referencing documents - scholarly papers, legal opinions, policy briefs - encode the topology of reference but not its stance. The standard representation collapses a rich evaluative relation into an untyped edge, losing the very content that supports community-level queries about how one document is received by another. We propose the claim network: a representational pattern in which each cross-document reference is reified as a typed claim, carrying source, target, claim text, and a four-class stance label grounded in the citation-intent literature. We give a construction pipeline applicable to any corpus of scholarly inter-referencing documents and instantiate it on a corpus of 127 papers in 3D point cloud semantic segmentation, producing a network of 8,260 typed claims. Three downstream task families demonstrate what the network enables: retrieval signal augmentation, aggregated-stance summarisation, and topological analytics. Head-to-head evaluation against standard Retrieval-Augmented Generation (RAG) baselines shows that the gain over flat retrieval is the gain from the right intermediate representation rather than the wrong one.

Story rewriting aims to adapt existing narratives to diverse reader preferences while preserving plot consistency and narrative coherence. Unlike conventional work on style transfer, we argue that effective story rewriting demands context-aware narrative enrichment beyond surface-level stylistic adaptation. Our pilot human study shows that style adaptation alone provides only marginal gains in reader satisfaction (2.3%), while context-enhanced rewriting substantially improves user preference alignment (24.5%). Motivated by this, we introduce STORYLENSBENCH, a large-scale benchmark for preference-aligned story rewriting, comprising structured story books, multi-dimensional reader preference profiles, and ranked context-aware rewritten stories. Building on this benchmark, we propose STORYLENSEVAL, a reward model for estimating reader satisfaction over rewritten stories, and STORYLENSWRITER, a two-stage rewriting model combining supervised fine-tuning with GRPO-based reinforcement learning. We further establish a comprehensive evaluation framework covering fidelity, coherence, and reader satisfaction. Experimental results demonstrate that STORYLENSWRITER consistently outperforms strong generation and personalization baselines, highlighting the importance of context-aware narrative enrichment for personalized story rewriting.

Audio large language models (Audio LLMs) demonstrate strong performance on speech understanding tasks, yet their ability to understand paralinguistic information remains limited. To systematically quantify this issue, we introduce VoxParadox, an adversarial benchmark with 2,000 verified examples, spanning 10 paralinguistic tasks, created with controlled speech synthesis to intentionally mismatch transcript claims and speaking style, enabling direct measurement of speech paralinguistic understanding. Evaluation of a diverse set of Audio LLMs reveals consistently low accuracy on acoustic ground truth and a strong tendency to follow language-implied (incorrect) answers. To understand the cause of this gap, we perform layer-wise probing and find that (i) paralinguistic cues can degrade in deeper encoder layers and at the encoder--LLM interface, and (ii) even when such cues are available in audio tokens, the language model frequently ignores them. To address these problems, we propose Prompt-Conditioned Layer Mixer (PCLM), which adaptively combines information from multiple audio layers based on the input prompt, and pair it with Direct Preference Optimization (DPO) to explicitly prefer acoustically supported options over language-implied alternatives. These methods substantially improve Audio LLM paralinguistic understanding, improving Audio Flamingo 3 from 17.40% to 65.20% on VoxParadox, and from 37.74% to 54.78% on MMSU paralinguistic subset. Our project page is available at https://voxparadox.github.io/.

Recent work has shown that Vision-Language Models (VLMs) used for optical character recognition (OCR) can generate plausible but visually unsupported text, suggesting reliance on language priors. Comparing open-weight VLMs with traditional OCR baselines on low-resource Ancient Greek critical editions, we show that VLM errors often remain fluent even when wrong, producing plausible Greek substitutions where traditional engines produce local recognition noise. To analyze visual evidence during decoding, we introduce controlled image perturbations and token-level grounding measures based on conditional versus image-free decoding distributions. Under character-level perturbations, VLMs diverge sharply from the perturbed ground truth while traditional OCR remains comparatively faithful; however, token-level analysis shows that prior reliance is model-specific: in an OCR-specialist model, fluent lexical errors are produced with little reliance on the image, whereas general-purpose VLMs remain conditioned on the visual input even when wrong. Decode-time interventions fail to reliably restore grounding, while post-OCR language-model correction improves several systems only by repairing text after generation. Our results extend prior evidence of OCR language-prior reliance to low-resource historical documents and a broader set of models, showing that fluent output is not necessarily visually grounded and motivating interpretability-driven evaluation beyond aggregate accuracy.

Retrieval-augmented generation (RAG) systems can respond incorrectly even when the correct passage was retrieved. The model must still read the retrieved passages and identify which one contains the answer among others that look relevant. This passage-reading model is called the reader. Does it fail simply because the context is longer or because the other passages genuinely compete with the correct one? We introduce and demonstrate a matched-control protocol for RAG reading: we keep the number and length of passages fixed, but replace hard competitors with less competitive real passages. We apply this control across two compact open models on SQuAD. This replacement partially restores performance, with the strongest effects on F1 and answer inclusion. For Phi-2, this recovers +6.0 EM points, +7.0 answer-inclusion points, and +0.057 F1. For Qwen2.5-1.5B, it recovers +4.5 EM points, +9.0 answer-inclusion points, and +0.068 F1. To track how performance changes as competitors accumulate, we also report retention curves and summarize them with a right-censored half-life when the curves do not cross half-retention. Together, these results show the protocol isolates a competition effect distinct from context length, though the effect is clearer for F1 and answer inclusion than for exact match, and also varies with snippet length.

Recent advances in large language models (LLMs) have made it increasingly difficult to distinguish human-written text from AI-generated content. Many existing detectors train supervised neural classifiers that achieve strong in-distribution performance but are often opaque and can degrade substantially under distribution shift. We present READER, a reasoning-enhanced AI text detector that outputs both a human/AI label and a structured rationale describing the evidence for its decision. A key component of our approach is READ, a curated supervision set of rationales and verdicts. We fine-tune an LLM on READ to build READER, which reasons before detecting at inference time. Despite having only 1.5B parameters, READER consistently outperforms existing detectors as well as prompted, high-capacity LLM baselines (GPT-5.2, Gemini-3-Pro, and DeepSeek-V3.2), which are 100 to 1000 times larger in scale.

A key question in psycholinguistics is how inferences about the meaning of linguistic input unfold incrementally a comprehender's mind. In this work, we study reading dynamics for ``noisy-channel garden-path'' sentences, which temporarily appear well-formed but feature late-appearing violations of expectation that can be resolved not by inferring an alternative syntactic structure, but by inferring the presence of an error. We find evidence for targeted regressions -- eye movements towards regions that are promising loci of possible errors in light of later-arriving information, showing patterns consistent with the posterior inferences of a model of noisy-channel processing with reanalysis. We discuss the implications of these findings for theories of noisy-channel language comprehension and information-theoretic explanations of reading dynamics.

In recent advancements within speech processing, converting read speech to conversational speech has gained significant attention. The primary challenge in this domain is maintaining naturalness and intelligibility while minimizing computational overhead for real-time applications. Traditional read speech often lacks the nuanced prosodic variation essential for natural conversational interactions, posing challenges for applications in virtual assistants, customer service, and language learning tools. This paper introduces a novel approach, Prosodic Adjustment with Conversational Context (PACC), aimed at converting read speech into natural conversational speech used in various modern applications. PACC utilizes advanced deep neural networks to analyze and modify prosodic features such as intonation, stress, and rhythm. Unlike conventional methods, our approach uses High-Fidelity Generative Adversarial Networks (HiFi-GAN) for speech synthesis. Our experimental results demonstrate significant improvements in speech conversion, enhancing naturalness and achieving better model accuracy with additional training on speech datasets. This research establishes new benchmarks in speech conversion tasks and Mean Opinion Score (MOS) evaluation for testing model accuracy, and we show that our approach can be successfully extended to other speech conversion applications.

As generative AI becomes increasingly integrated into journalism, designing effective AI-use disclosures that inform readers without imposing unnecessary burden is a key challenge. While prior research has primarily focused on trust and credibility, the impact of disclosures on readers' attentional and cognitive load remains underexplored. To address this gap, we conducted a $3\times2\times2$ mixed factorial study manipulating the level of AI-use disclosure detail (none, one-line, detailed), news type (politics, lifestyle), and role of AI (editing, partial content generation), measuring load via NASA-TLX and eye-tracking. Our results reveal a significant attentional cost: one-line disclosures resulted in significantly higher fixation durations and saccade counts, particularly for AI-edited content. Detailed disclosures did not impose additional burden. Drawing on Information-Gap Theory, we argue that brief labels may trigger increased visual scrutiny by alerting readers to AI use without providing enough information. NASA-TLX scores and pupil diameter showed no significant differences across conditions, suggesting that AI-use disclosures do not impose cognitive burden regardless of the detail level. Interview insights contextualize these findings and reveal a strong preference for detailed or ``detail-on-demand'' designs. Our findings inform the design of gaze-informed adaptive disclosure interfaces that dynamically adjust transparency levels based on readers' attentional patterns and news context.

What does a book look like to a quantum computer? This paper takes eight classical works of the Renaissance and its late-antique inheritance -- from Augustine to Galileo -- and runs each through a neutral-atom quantum processor. The bridge is graphs: each textual unit becomes an atom, and graph edges are physical blockade constraints for engineered exact unit-disk designs, or a 2D approximation to the semantic graph for natural texts. Three contributions follow. First, we introduce rigidity rho, a metric for how unique a book's structural backbone is -- distinguishing Marguerite de Navarre's Heptameron (rigid, twelve-nouvelle hard core) from Boethius (fully fungible, every chapter substitutable). Second, we invert the pipeline: rather than extracting a graph from existing prose, we pick a target graph the hardware encodes natively, and write a book whose structure matches it. The twenty-nine texts written this way, collected under the name QOuLiPo, extend the OuLiPo tradition to graph-topological constraints and, together with the eight natural texts, form a benchmark distribution against which neutral-atom hardware can be tracked as it scales. Third, we run both natural and engineered texts on Pasqal's FRESNEL processor up to one hundred atoms; engineered texts reach high approximation ratios, the cleanest instances returning the exact backbone. A cloud-accessible quantum machine plus an agentic coding environment now lets a single investigator run this pipeline end-to-end. What is reported is an application layer, not a speedup -- humanistic instances ready to load onto neutral-atom processors as they scale, already complementing classical text analysis. The Digital Humanities community has a stake in building familiarity with this hardware now: the engineered-corpus design choices made today fix the benchmark distribution future hardware will be measured against.

The discovery of novel methodologies for emerging problems is a continuing cycle in ML, often driven by the migration of techniques across domains. Building on this observation, we ask whether current LLM ideation systems benefit from targeted cross-domain retrieval or simply from exposure to diverse mechanisms. We study this question through PaperGym, a three-stage pipeline: (1) tool-augmented seed extraction via read, grep, and bash over an isolated paper environment, (2) cross-domain seed retrieval via paraphrasing across seven ML domains, and (3) method synthesis from retrieved seeds, each scored by rubric-based judges. Tool-augmented extraction improves specificity, and paraphrase-based retrieval broadens domain coverage. In synthesis, cross-domain retrieval receives more pairwise novelty wins than no-retrieval and same-domain baselines, but shows no significant difference from a random diverse-seed control. These findings suggest LLM ideation systems benefit from diverse seed exposure, but do not yet reliably exploit the semantic reason particular seeds were retrieved. We release the seed library, rubric prompts, and run scripts at https://github.com/yunjoochoi/PaperGym

Capability distillation applies knowledge distillation to selected model capabilities, aiming to compress a large language model (LLM) into a smaller one while preserving the abilities needed for a downstream task. However, most existing methods treat capabilities as independent training targets and overlook how improving one capability can reshape the student's broader capability profile, especially when multiple abilities jointly determine task success. We study capability distillation under a fixed token budget and identify two consistent patterns: distillation induces systematic, budget-dependent cross-capability transfer, and additional budget often brings limited task-relevant gains while sometimes degrading other useful abilities. Building on these insights, we propose ReAD, a Reinforcement-guided cApability Distillation framework that explicitly accounts for capability interdependence. ReAD first infers task-essential capabilities, then generates capability-targeted supervision on the fly, and finally uses an uncertainty-aware contextual bandit to adaptively allocate the distillation budget based on expected utility gains. Extensive experiments show that ReAD improves downstream utility under the same token budget while reducing harmful spillover and wasted distillation effort compared to strong baselines. Our code is publicly available at https://github.com/LabRAI/ReAD.

This paper presents the ReXCL tool, which automates the extraction and classification processes in requirements engineering, enhancing the software development life-cycle. The tool features two main modules: Extraction, which processes raw requirement documents into a predefined schema using heuristics and predictive modeling, and Classification, which assigns class labels to requirements using adaptive fine-tuning of encoder-based models. The final output can be exported to external requirement engineering tools. Performance evaluations indicate that ReXCL significantly improves efficiency and accuracy in managing requirements, marking a novel approach to automating the schematization of semi-structured requirement documents.

Accurate text recognition in low-light environments is essential for intelligent systems in applications ranging from autonomous vehicles to smart surveillance. However, challenges such as poor illumination and noise interference remain underexplored. To address this gap, we introduce LSTR, a large-scale Low-light Scene Text Recognition dataset comprising 11,273 low-light images generated from well-lit datasets (ICDAR2015, IIIT5K, and WordArt), along with ESTR, which includes 60 real nighttime street-scene images in English and Spanish for exclusive evaluation. We explore two solution strategies: (1) employing Optical Character Recognition (OCR) models with fine-tuning and LoRA-based fine-tuning and (2) a joint training strategy that integrates a low-light image enhancement (LLIE) module with an OCR model. In particular, we propose a novel re-render LLIE (RLLIE) module, which demonstrates improved performance on real-world data. Through extensive experimentation, we analyze various training strategies and address a key research question: \emph{How bright is bright enough for effective scene text recognition?} Our results indicate that standalone LLIE or OCR models perform inadequately under low-light conditions, highlighting the advantages of specialized, jointly trained text-centric approaches. Additionally, we provide a comprehensive benchmark to support future research in robust low-light scene text recognition. https://huggingface.co/datasets/lumimusta/Low-light_Scene_Text_Dataset.