The Limits of Learning from Pictures and Text: Vision-Language Models and Embodied Scene Understanding

What information is sufficient to learn the full richness of human scene understanding? The distributional hypothesis holds that the statistical co-occurrence of language and images captures the conceptual knowledge underlying visual cognition. Vision-language models (VLMs) are trained on massive paired text-image corpora but lack embodied experience, making them an ideal test of the distributional hypothesis. We report two experiments comparing descriptions generated by 18 VLMs to those of over 2000 human observers across 15 high-level scene understanding tasks, spanning general knowledge, affordances, sensory experiences, affective responses, and future prediction. Because many tasks lack ground truth answers, we developed a Human-Calibrated Cosine Distance (HCD) metric that measures VLM output similarity to the distribution of human responses, scaled by within-human variability. In Experiment 1, VLMs approached human-level performance on general knowledge tasks, but showed a robust deficit for affordance tasks that resisted prompt engineering and did not improve with newer model releases. In Experiment 2, we tested six mechanistic hypotheses for explaining this affordance gap, finding that the deficit was structural rather than stylistic and was not resolved by providing explicit spatial information. Corpus analyses revealed that image captioning datasets contain sparse agent-addressed affordance language, consistent with Gricean accounts of why embodied knowledge may be systematically underrepresented in language. Together, these findings suggest that distributional learning from images and text is insufficient for affordance-based scene understanding, implying that some dimensions of human visual cognition may require the kind of agent-centered, three-dimensional experience that no photograph or caption can encode.

EnTaCs: Analyzing the Relationship Between Sentiment and Language Choice in English-Tamil Code-Switching

This paper investigates the relationship between utterance sentiment and language choice in English-Tamil code-switched text, using methods from machine learning and statistical modelling. We apply a fine-tuned XLM-RoBERTa model for token-level language identification on 35,650 romanized YouTube comments from the DravidianCodeMix dataset, producing per-utterance measurements of English proportion and language switch frequency. Linear regression analysis reveals that positive utterances exhibit significantly greater English proportion (34.3%) than negative utterances (24.8%), and mixed-sentiment utterances show the highest language switch frequency when controlling for utterance length. These findings support the hypothesis that emotional content demonstrably influences language choice in multilingual code-switching settings, due to socio-linguistic associations of prestige and identity with embedded and matrix languages.

Generation Is Compression: Zero-Shot Video Coding via Stochastic Rectified Flow

Existing generative video compression methods use generative models only as post-hoc reconstruction modules atop conventional codecs. We propose \emph{Generative Video Codec} (GVC), a zero-shot framework that turns a pretrained video generative model into the codec itself: the transmitted bitstream directly specifies the generative decoding trajectory, with no retraining required. To enable this, we convert the deterministic rectified-flow ODE of modern video foundation models into an equivalent SDE at inference time, unlocking per-step stochastic injection points for codebook-driven compression. Building on this unified backbone, we instantiate three complementary conditioning strategies -- \emph{Image-to-Video} (I2V) with adaptive tail-frame atom allocation, \emph{Text-to-Video} (T2V) operating at near-zero side information as a pure generative prior, and \emph{First-Last-Frame-to-Video} (FLF2V) with boundary-sharing GOP chaining for dual-anchor temporal control. Together, these variants span a principled trade-off space between spatial fidelity, temporal coherence, and compression efficiency. Experiments on standard benchmarks show that GVC achieves high-quality reconstruction below 0.002\,bpp while supporting flexible bitrate control through a single hyperparameter.

CADSmith: Multi-Agent CAD Generation with Programmatic Geometric Validation

Existing methods for text-to-CAD generation either operate in a single pass with no geometric verification or rely on lossy visual feedback that cannot resolve dimensional errors. We present CADSmith, a multi-agent pipeline that generates CadQuery code from natural language. It then undergoes an iterative refinement process through two nested correction loops: an inner loop that resolves execution errors and an outer loop grounded in programmatic geometric validation. The outer loop combines exact measurements from the OpenCASCADE kernel (bounding box dimensions, volume, solid validity) with holistic visual assessment from an independent vision-language model Judge. This provides both the numerical precision and the high-level shape awareness needed to converge on the correct geometry. The system uses retrieval-augmented generation over API documentation rather than fine-tuning, maintaining a current database as the underlying CAD library evolves. We evaluate on a custom benchmark of 100 prompts in three difficulty tiers (T1 through T3) with three ablation configurations. Against a zero-shot baseline, CADSmith achieves a 100% execution rate (up from 95%), improves the median F1 score from 0.9707 to 0.9846, the median IoU from 0.8085 to 0.9629, and reduces the mean Chamfer Distance from 28.37 to 0.74, demonstrating that closed-loop refinement with programmatic geometric feedback substantially improves the quality and reliability of LLM-generated CAD models.

Can AI Models Direct Each Other? Organizational Structure as a Probe into Training Limitations

Can an expensive AI model effectively direct a cheap one to solve software engineering tasks? We study this question by introducing ManagerWorker, a two-agent pipeline where an expensive "manager" model (text-only, no code execution) analyzes issues, dispatches exploration tasks, and reviews implementations, while a cheap "worker" model (with full repo access) executes code changes. We evaluate on 200 instances from SWE-bench Lite across five configurations that vary the manager-worker relationship, pipeline complexity, and model pairing. Our findings reveal both the promise and the limits of multi-agent direction: (1) a strong manager directing a weak worker (62%) matches a strong single agent (60%) at a fraction of the strong-model token usage, showing that expensive reasoning can substitute for expensive execution; (2) a weak manager directing a weak worker (42%) performs worse than the weak agent alone (44%), demonstrating that the directing relationship requires a genuine capability gap--structure without substance is pure overhead; (3) the manager's value lies in directing, not merely reviewing--a minimal review-only loop adds just 2pp over the baseline, while structured exploration and planning add 11pp, showing that active direction is what makes the capability gap productive; and (4) these behaviors trace to a single root cause: current models are trained as monolithic agents, and splitting them into director/worker roles fights their training distribution. The pipeline succeeds by designing around this mismatch--keeping each model close to its trained mode (text generation for the manager, tool use for the worker) and externalizing organizational structure to code. This diagnosis points to concrete training gaps: delegation, scoped execution, and mode switching are skills absent from current training data.

Automating Clinical Information Retrieval from Finnish Electronic Health Records Using Large Language Models

Clinicians often need to retrieve patient-specific information from electronic health records (EHRs), a task that is time-consuming and error-prone. We present a locally deployable Clinical Contextual Question Answering (CCQA) framework that answers clinical questions directly from EHRs without external data transfer. Open-source large language models (LLMs) ranging from 4B to 70B parameters were benchmarked under fully offline conditions using 1,664 expert-annotated question-answer pairs derived from records of 183 patients. The dataset consisted predominantly of Finnish clinical text. In free-text generation, Llama-3.1-70B achieved 95.3% accuracy and 97.3% consistency across semantically equivalent question variants, while the smaller Qwen3-30B-A3B-2507 model achieved comparable performance. In a multiple-choice setting, models showed similar accuracy but variable calibration. Low-precision quantization (4-bit and 8-bit) preserved predictive performance while reducing GPU memory requirements and improving deployment feasibility. Clinical evaluation identified clinically significant errors in 2.9% of outputs, and semantically equivalent questions occasionally yielded discordant responses, including instances where one formulation was correct and the other contained a clinically significant error (0.96% of cases). These findings demonstrate that locally hosted open-source LLMs can accurately retrieve patient-specific information from EHRs using natural-language queries, while highlighting the need for validation and human oversight in clinical deployment.

Why Models Know But Don't Say: Chain-of-Thought Faithfulness Divergence Between Thinking Tokens and Answers in Open-Weight Reasoning Models

Extended-thinking models expose a second text-generation channel ("thinking tokens") alongside the user-visible answer. This study examines 12 open-weight reasoning models on MMLU and GPQA questions paired with misleading hints. Among the 10,506 cases where models actually followed the hint (choosing the hint's target over the ground truth), each case is classified by whether the model acknowledges the hint in its thinking tokens, its answer text, both, or neither. In 55.4% of these cases the model's thinking tokens contain hint-related keywords that the visible answer omits entirely, a pattern termed *thinking-answer divergence*. The reverse (answer-only acknowledgment) is near-zero (0.5%), confirming that the asymmetry is directional. Hint type shapes the pattern sharply: sycophancy is the most *transparent* hint, with 58.8% of sycophancy-influenced cases acknowledging the professor's authority in both channels, while consistency (72.2%) and unethical (62.7%) hints are dominated by thinking-only acknowledgment. Models also vary widely, from near-total divergence (Step-3.5-Flash: 94.7%) to relative transparency (Qwen3.5-27B: 19.6%). These results show that answer-text-only monitoring misses more than half of all hint-influenced reasoning and that thinking-token access, while necessary, still leaves 11.8% of cases with no verbalized acknowledgment in either channel.

A Formal Framework for Uncertainty Analysis of Text Generation with Large Language Models

The generation of texts using Large Language Models (LLMs) is inherently uncertain, with sources of uncertainty being not only the generation of texts, but also the prompt used and the downstream interpretation. Within this work, we provide a formal framework for the measurement of uncertainty that takes these different aspects into account. Our framework models prompting, generation, and interpretation as interconnected autoregressive processes that can be combined into a single sampling tree. We introduce filters and objective functions to describe how different aspects of uncertainty can be expressed over the sampling tree and demonstrate how to express existing approaches towards uncertainty through these functions. With our framework we show not only how different methods are formally related and can be reduced to a common core, but also point out additional aspects of uncertainty that have not yet been studied.

HINT: Composed Image Retrieval with Dual-path Compositional Contextualized Network

Composed Image Retrieval (CIR) is a challenging image retrieval paradigm. It aims to retrieve target images from large-scale image databases that are consistent with the modification semantics, based on a multimodal query composed of a reference image and modification text. Although existing methods have made significant progress in cross-modal alignment and feature fusion, a key flaw remains: the neglect of contextual information in discriminating matching samples. However, addressing this limitation is not an easy task due to two challenges: 1) implicit dependencies and 2) the lack of a differential amplification mechanism. To address these challenges, we propose a dual-patH composItional coNtextualized neTwork (HINT), which can perform contextualized encoding and amplify the similarity differences between matching and non-matching samples, thus improving the upper performance of CIR models in complex scenarios. Our HINT model achieves optimal performance on all metrics across two CIR benchmark datasets, demonstrating the superiority of our HINT model. Codes are available at https://github.com/zh-mingyu/HINT.

Verify Claimed Text-to-Image Models via Boundary-Aware Prompt Optimization

As Text-to-Image (T2I) generation becomes widespread, third-party platforms increasingly integrate multiple model APIs for convenient image creation. However, false claims of using official models can mislead users and harm model owners' reputations, making model verification essential to confirm whether an API's underlying model matches its claim. Existing methods address this by using verification prompts generated by official model owners, but the generation relies on multiple reference models for optimization, leading to high computational cost and sensitivity to model selection. To address this problem, we propose a reference-free T2I model verification method called Boundary-aware Prompt Optimization (BPO). It directly explores the intrinsic characteristics of the target model. The key insight is that although different T2I models produce similar outputs for normal prompts, their semantic boundaries in the embedding space (transition zones between two concepts such as "corgi" and "bagel") are distinct. Prompts near these boundaries generate unstable outputs (e.g., sometimes a corgi and sometimes a bagel) on the target model but remain stable on other models. By identifying such boundary-adjacent prompts, BPO captures model-specific behaviors that serve as reliable verification cues for distinguishing T2I models. Experiments on five T2I models and four baselines demonstrate that BPO achieves superior verification accuracy.