As large language models become increasingly capable, there is growing concern that they may develop reasoning processes that are encoded or hidden from human oversight. To investigate whether current interpretability techniques can penetrate such encoded reasoning, we construct a controlled testbed by fine-tuning a reasoning model (DeepSeek-R1-Distill-Llama-70B) to perform chain-of-thought reasoning in ROT-13 encryption while maintaining intelligible English outputs. We evaluate mechanistic interpretability methods--in particular, logit lens analysis--on their ability to decode the model's hidden reasoning process using only internal activations. We show that logit lens can effectively translate encoded reasoning, with accuracy peaking in intermediate-to-late layers. Finally, we develop a fully unsupervised decoding pipeline that combines logit lens with automated paraphrasing, achieving substantial accuracy in reconstructing complete reasoning transcripts from internal model representations. These findings suggest that current mechanistic interpretability techniques may be more robust to simple forms of encoded reasoning than previously understood. Our work provides an initial framework for evaluating interpretability methods against models that reason in non-human-readable formats, contributing to the broader challenge of maintaining oversight over increasingly capable AI systems.

We introduce, a large-scale synthetic benchmark of 15,045 university-level physics problems (90/10% train/test split). Each problem is fully parameterized, supporting an effectively infinite range of input configurations, and is accompanied by structured, step-by-step reasoning and executable Python code that produces the ground-truth solution for any parameter set. The benchmark contains three question types: MC-Symbolic (multiple-choice with symbolic options), MC-Numerical (multiple-choice with numerical options), and free-form (open-ended responses). These diverse formats test complementary reasoning skills. By leveraging the dynamic, code-driven nature of the benchmark, we introduce three novel evaluation metrics in addition to standard accuracy: Consistency Score, Failure Rate, and Confusion Rate, that quantify variability and uncertainty across problem variants. Experiments with state-of-the-art instruction-tuned language models reveal both strengths and limitations in scientific reasoning, positioning SymPyBench as a foundation for developing more robust and interpretable reasoning systems

Evaluating vision-language models (VLMs) in scientific domains like mathematics and physics poses unique challenges that go far beyond predicting final answers. These domains demand conceptual understanding, symbolic reasoning, and adherence to formal laws, requirements that most existing benchmarks fail to address. In particular, current datasets tend to be static, lacking intermediate reasoning steps, robustness to variations, or mechanisms for verifying scientific correctness. To address these limitations, we introduce PRiSM, a synthetic, fully dynamic, and multimodal benchmark for evaluating scientific reasoning via grounded Python code. PRiSM includes over 24,750 university-level physics and math problems, and it leverages our scalable agent-based pipeline, PrismAgent, to generate well-structured problem instances. Each problem contains dynamic textual and visual input, a generated figure, alongside rich structured outputs: executable Python code for ground truth generation and verification, and detailed step-by-step reasoning. The dynamic nature and Python-powered automated ground truth generation of our benchmark allow for fine-grained experimental auditing of multimodal VLMs, revealing failure modes, uncertainty behaviors, and limitations in scientific reasoning. To this end, we propose five targeted evaluation tasks covering generalization, symbolic program synthesis, perturbation robustness, reasoning correction, and ambiguity resolution. Through comprehensive evaluation of existing VLMs, we highlight their limitations and showcase how PRiSM enables deeper insights into their scientific reasoning capabilities.

Vision-Language Models (VLMs) remain limited in spatial reasoning tasks that require multi-view understanding and embodied perspective shifts. Recent approaches such as MindJourney attempt to mitigate this gap through test-time scaling where a world model imagines action-conditioned trajectories and a heuristic verifier selects helpful views from such trajectories. In this work, we systematically examine how such test-time verifiers behave across benchmarks, uncovering both their promise and their pitfalls. Our uncertainty-based analyses show that MindJourney's verifier provides little meaningful calibration, and that random scoring often reduces answer entropy equally well, thus exposing systematic action biases and unreliable reward signals. To mitigate these, we introduce a Verification through Spatial Assertions (ViSA) framework that grounds the test-time reward in verifiable, frame-anchored micro-claims. This principled verifier consistently improves spatial reasoning on the SAT-Real benchmark and corrects trajectory-selection biases through more balanced exploratory behavior. However, on the challenging MMSI-Bench, none of the verifiers, including ours, achieve consistent scaling, suggesting that the current world models form an information bottleneck where imagined views fail to enrich fine-grained reasoning. Together, these findings chart the bad, good, and ugly aspects of test-time verification for world-model-based reasoning. Our code is available at https://github.com/chandar-lab/visa-for-mindjourney.

Influential critiques argue that Large Language Models (LLMs) are a dead end for AGI: "mere pattern matchers" structurally incapable of reasoning or planning. We argue this conclusion misidentifies the bottleneck: it confuses the ocean with the net. Pattern repositories are the necessary System-1 substrate; the missing component is a System-2 coordination layer that selects, constrains, and binds these patterns. We formalize this layer via UCCT, a theory of semantic anchoring that models reasoning as a phase transition governed by effective support (rho_d), representational mismatch (d_r), and an adaptive anchoring budget (gamma log k). Under this lens, ungrounded generation is simply an unbaited retrieval of the substrate's maximum likelihood prior, while "reasoning" emerges when anchors shift the posterior toward goal-directed constraints. We translate UCCT into architecture with MACI, a coordination stack that implements baiting (behavior-modulated debate), filtering (Socratic judging), and persistence (transactional memory). By reframing common objections as testable coordination failures, we argue that the path to AGI runs through LLMs, not around them.

We introduce FieldSeer I, a geometry-aware world model that forecasts electromagnetic field dynamics from partial observations in 2-D TE waveguides. The model assimilates a short prefix of observed fields, conditions on a scalar source action and structure/material map, and generates closed-loop rollouts in the physical domain. Training in a symmetric-log domain ensures numerical stability. Evaluated on a reproducible FDTD benchmark (200 unique simulations, structure-wise split), FieldSeer I achieves higher suffix fidelity than GRU and deterministic baselines across three practical settings: (i) software-in-the-loop filtering (64x64, P=80->Q=80), (ii) offline single-file rollouts (80x140, P=240->Q=40), and (iii) offline multi-structure rollouts (80x140, P=180->Q=100). Crucially, it enables edit-after-prefix geometry modifications without re-assimilation. Results demonstrate that geometry-conditioned world models provide a practical path toward interactive digital twins for photonic design.

Recent advancements in large reasoning models have fueled growing interest in extending such capabilities to multimodal domains. However, despite notable progress in visual reasoning, the lack of transparent and reproducible data curation and training strategies remains a major barrier to scalable research. In this work, we introduce OpenMMReasoner, a fully transparent two-stage recipe for multimodal reasoning spanning supervised fine-tuning (SFT) and reinforcement learning (RL). In the SFT stage, we construct an 874K-sample cold-start dataset with rigorous step-by-step validation, providing a strong foundation for reasoning capabilities. The subsequent RL stage leverages a 74K-sample dataset across diverse domains to further sharpen and stabilize these abilities, resulting in a more robust and efficient learning process. Extensive evaluations demonstrate that our training recipe not only surpasses strong baselines but also highlights the critical role of data quality and training design in shaping multimodal reasoning performance. Notably, our method achieves a 11.6% improvement over the Qwen2.5-VL-7B-Instruct baseline across nine multimodal reasoning benchmarks, establishing a solid empirical foundation for future large-scale multimodal reasoning research. We open-sourced all our codes, pipeline, and data at https://github.com/EvolvingLMMs-Lab/OpenMMReasoner.

Retrosynthesis prediction aims to infer the reactant molecule based on a given product molecule, which is a fundamental task in chemical synthesis. However, existing models rely on static pattern-matching paradigm, which limits their ability to perform effective logic decision-making, leading to black-box decision-making. Building on this, we propose Retro-Expert, an interpretable retrosyn-thesis framework that performs collaborative reasoning by combining the complementary reasoning strengths of Large Language Models and specialized models via reinforcement learning. It outputs natural language explanations grounded in chemical logic through three components: (1) specialized models analyze the product to construct high-quality chemical decision space, (2) LLM-driven critical reasoning to generate predictions and corresponding interpretable reasoning path, and (3) reinforcement learning optimizing interpretable decision policy. Experiments show that Retro-Expert not only surpasses both LLM-based and specialized models across different metrics but also provides expert-aligned explanations that bridge the gap between AI predictions and actionable chemical insights.

Agentic AI systems, software agents with autonomy, decision-making ability, and adaptability, are increasingly used to execute complex tasks on behalf of organisations. Most such systems rely on Large Language Models (LLMs), whose broad semantic capabilities enable powerful language processing but lack explicit, institution-specific grounding. In enterprises, data rarely comes with an inspectable semantic layer, and constructing one typically requires labour-intensive "data archaeology": cleaning, modelling, and curating knowledge into ontologies, taxonomies, and other formal structures. At the same time, explainability methods such as saliency maps expose an "interpretability gap": they highlight what the model attends to but not why, leaving decision processes opaque. In this preprint, we present a case study, developed by Kaiasm and Avantra AI through their work with The Turing Way Practitioners Hub, a forum developed under the InnovateUK BridgeAI program. This study presents a collaborative human-AI approach to building an inspectable semantic layer for Agentic AI. AI agents first propose candidate knowledge structures from diverse data sources; domain experts then validate, correct, and extend these structures, with their feedback used to improve subsequent models. Authors show how this process captures tacit institutional knowledge, improves response quality and efficiency, and mitigates institutional amnesia. We argue for a shift from post-hoc explanation to justifiable Agentic AI, where decisions are grounded in explicit, inspectable evidence and reasoning accessible to both experts and non-specialists.

Self-adaptive systems (SASs) are designed to handle changes and uncertainties through a feedback loop with four core functionalities: monitoring, analyzing, planning, and execution. Recently, generative artificial intelligence (GenAI), especially the area of large language models, has shown impressive performance in data comprehension and logical reasoning. These capabilities are highly aligned with the functionalities required in SASs, suggesting a strong potential to employ GenAI to enhance SASs. However, the specific benefits and challenges of employing GenAI in SASs remain unclear. Yet, providing a comprehensive understanding of these benefits and challenges is complex due to several reasons: limited publications in the SAS field, the technological and application diversity within SASs, and the rapid evolution of GenAI technologies. To that end, this paper aims to provide researchers and practitioners a comprehensive snapshot that outlines the potential benefits and challenges of employing GenAI's within SAS. Specifically, we gather, filter, and analyze literature from four distinct research fields and organize them into two main categories to potential benefits: (i) enhancements to the autonomy of SASs centered around the specific functions of the MAPE-K feedback loop, and (ii) improvements in the interaction between humans and SASs within human-on-the-loop settings. From our study, we outline a research roadmap that highlights the challenges of integrating GenAI into SASs. The roadmap starts with outlining key research challenges that need to be tackled to exploit the potential for applying GenAI in the field of SAS. The roadmap concludes with a practical reflection, elaborating on current shortcomings of GenAI and proposing possible mitigation strategies.