LLM 理论进展

Deception plays a central role in Intelligence operations, yet it remains difficult to analyse systematically without expert knowledge of reasoning patterns and cognitive manipulation. In computational argumentation, for instance, no scheme-level ground-truth corpora currently exist to support statistical validation. In this paper, we address this by introducing a set of ten argument schemes designed to model distinct forms of deception, each accompanied by structured premises and critical questions. In doing so, we introduce the first dedicated library of argumentation schemes specifically designed for deception analysis, providing a structured foundation for systematically modelling and analysing deception in narrative text. We then present DeceptionAnalyser, a browser-based tool that implements these schemes through a two-stage methodology combining LLM-based premise extraction with critical-question-driven evaluation. Our aim is to provide a conceptual and methodological foundation for analysing deceptive reasoning in narrative text. This is precisely what we address in this paper by demonstrating how structured argumentation theory and AI-assisted analysis can support transparent, explainable assessments of potential deception. Because the schemes are designed to flag claims for scrutiny rather than to output a deception verdict, we do not benchmark classification accuracy; instead, we assess the reliability of the methodology by measuring the consistency of the tool's premise and conclusion assessments across ten contemporary large language models and repeated runs. We find that scheme detection is highly stable for clear-cut deception and degrades gracefully, in interpretable ways, on more ambiguous intelligence-style narratives.

The effectiveness of agent memory ultimately depends on whether the underlying LLM gives each memory in context an appropriate degree of influence over its response. Yet this capability has remained largely overlooked. To assess this capability, we introduce MemCalib, a benchmark grounded in realistic memory-system scenarios for evaluating memory use and advancing optimization algorithms. Results on the MemCalib test set reveal that frontier open- and closed-source models struggle to use memory appropriately. They frequently over-use or under-use memory rather than matching each proposition's actual use to its target level, leading to biased, low-quality responses. Experiments with common post-training algorithms, including group relative policy optimization and on-policy self-distillation, further reveal a clear directional skew: trained models improve in one direction while deteriorating in the other. We therefore propose MemCalib-RL, an ordered bidirectional counterfactual credit-assignment algorithm that separates over- and under-use signals and localizes their credit to response tokens through exact atom ablation. Results across model families and scales (Qwen3-8B, Ministral-3-8B-Instruct, and Qwen3.5-35B-A3B) show that MemCalib-RL achieves the best overall performance while better balancing over-use and under-use, with gains generalizing beyond MemCalib in external benchmark evaluation. Further experiments support its design choices and robustness and provide insight into its training dynamics.

Speculative decoding losslessly accelerates large language model inference by having a lightweight draft model predict future tokens for verification by the target model. Recent block diffusion drafters further reduce drafting latency by predicting multiple tokens in parallel. However, existing block drafters project target hidden states at every input position into a separate drafter-side KV cache, incurring per-request memory and KV-write overhead that grow with concurrency; directly reusing target KVs in place removes this cache but fails to sustain draft quality throughout the block. We propose a hybrid target-context injection method that complements direct target KV reuse with target hidden states only at the last input position, requiring no separate drafter-side KV cache. Building on this design, we propose H-Spec, a hybrid Mamba-attention parallel drafter that consumes the two target-context sources through complementary modules. Mamba modules are initialized with projected last-token target hidden states, while attention modules reuse target KVs in place. Despite its recurrent formulation, Mamba's parallel scan allows H-Spec to preserve block-parallel drafting. Across three target models and diverse tasks, H-Spec improves over the best baseline by 5.0--13.3% in mean accepted length and 5.3--12.6% in batch-size-1 inter-token latency speedup. Under concurrent serving, H-Spec consistently achieves higher throughput while maintaining lower KV cache utilization than baselines across evaluated concurrency levels.

Simulating student personas with large language models (LLMs) enables scalable evaluation of educational systems. However, behavioral drift, a progressive decline in persona consistency, can emerge over extended conversations, limiting the validity of such simulations. We evaluate five prompt-level mechanisms using separate monitoring and intervention pipelines. Across 1,200 28-turn conversations spanning four LLMs and two ADHD persona intensities, we varied when to intervene (static vs. adaptive) and what to inject (reinjection vs. reflective reminder), plus a novel adaptive condition in which a monitor generates behavior-specific instructions. Relative to no intervention, reinjection reduced the modeled rate of LLM-rated drift by 35--38%, reflective reminders by 22--27%, and behavior-specific instruction by 87%. None eliminated drift. We found no evidence that adaptive timing outperformed static scheduling. Monitoring therefore appears more useful for deciding what to correct than when to intervene, although behavior-specific instruction requires component-level testing.

Looped Language Models (LoopLMs) perform "latent reasoning" by recursively refining internal latent representations with shared weights, offering a more effective alternative to explicit verbal reasoning. Despite their effectiveness, we find that LoopLMs remain prone to loop instability: unstable refinement across iterations can produce localized uncertain "hard" tokens associated with reasoning errors. To address this, we propose LoopCD, loop-wise contrastive decoding that enhances the reasoning performance of LoopLMs by intervening on these tokens at inference time. Specifically, we exploit the internal dynamics of LoopLMs and contrast the logits from earlier iterations with logits from the last refined iteration to form the final sampling distribution. We find that this strategy is highly efficient, introducing only negligible inference overhead and requiring no additional training, while effectively improving reasoning performance by naturally refining reasoning-critical hard tokens. Extensive experiments show that our method improves the performance of recent representative LoopLMs across various reasoning tasks.

Long-context language models interface with external knowledge through raw natural language. In retrieval-augmented systems, this creates a persistent index-payload schism: dense vectors enable searchable routing, but models must re-ingest lengthy text payloads for reasoning at O(N^2) attention cost. Existing compression methods further produce private states tied to specific architectures. We introduce Machine-Interpretable Information (MII), the first agent-to-agent (A2A) document-to-state protocol. A dual-timescale state-space Writer compiles documents into a canonical, fixed-bandwidth state (56 tokens), and a lightweight Translator maps it into any frozen Reader's embedding space, reducing query-time cost to O(K). The resulting .mii artifact unifies Retrieval (searchable geometry), Reasoning (global memory), and Reconstruction (grounded details) in a single transferable medium. We demonstrate strong cross-model interoperability across heterogeneous LLMs (e.g., Llama, Qwen, Mistral) — despite the Writer using a legacy GPT-2 vocabulary, forcing genuine semantic translation rather than token-level memorization. Mechanistic probes reveal modular latent structure: entity representations can be causally traced and zero-shot transplanted between unrelated document states while remaining decodable. To address lexical reconstruction under fixed bandwidth, we propose Residual-MII, a cache hierarchy combining compiled global memory with sparse local evidence. On HotpotQA (7,405 queries), Residual-MII exceeds full-context Exact Match at approximately 7% of the attention FLOPs, suggesting a paradigm shift toward compiled, transferable neural document formats.

Large reasoning models (LRMs) incur high inference costs, often mitigated by efficiency techniques like quantization and pruning. However, the impact of these techniques on model adversarial robustness remains largely unexplored. This study provides the first comprehensive analysis of the interplay between efficiency, jailbreak vulnerability, and reasoning in LRMs. We find that while efficiency methods seemingly reduce the success rate of jailbreak attacks, this improvement is often superficial. It largely arises from degraded reasoning capabilities leading to "attempted but failed" malicious responses, rather than an increase in genuine alignment. Mechanistic analysis of representational drift confirms this, revealing a strict coupling between reasoning capability loss and the model's inability to maintain malicious semantic trajectories. Additionally, we identify quantization with pruning as the optimal strategy to balance efficiency and robustness. These findings clarify the distinction between true safety alignment and capability-induced failure, providing an empirical foundation for LRM deployment.

Large audio-language models (LALMs) are increasingly used for a broader range of audio reasoning tasks. These models typically incorporate audio representations into a large language model (LLM) backbone to enable multimodal reasoning. Recent test-time reinforcement learning (TTRL) methods further improve LLM reasoning capability by leveraging unlabelled test data after pre-training. However, the importance of the perceptual capability of LALMs remains underexplored, particularly how much acoustic evidence is integrated and relied upon during reasoning, and how this contributes to final task performance. This gap limits the development of effective post-training methods like TTRL for audio reasoning. In this work, we first analyse how audio information is integrated and utilised during reasoning process. We quantify layer-wise perceptual reliance and show that stronger acoustic reliance is associated with higher accuracy and a larger performance gain attributable to the audio input. Building on this, we propose Perception-Grounded TTRL (PG-TTRL), which aligns label-free test-time optimisation with perceptually grounded reasoning, encouraging the model to structure its reasoning more strongly on the audio input. Experiments across LALMs and benchmarks show that PG-TTRL consistently improves reasoning performance over both the base models and standard TTRL, showing the value of perceptual-grounding optimisation for test-time audio reasoning.

Memory systems are becoming a core component of LLM agents, but constructing and maintaining memory remains expensive because it relies on repeated calls to large proprietary language models. This cost creates a major barrier to deploying memory-enhanced agents at scale. In this paper, we present Pseudo Self-Distillation (PSD), a framework that enables small language models (SLMs) to construct hierarchical memory representations by distilling behavior from a strong black-box oracle through a multi-stage training pipeline. Standard distillation methods require access to teacher logits or hidden states, which closed models do not expose. Unlike conventional self-distillation settings, where supervision is derived from a model's own predictions, sampled rollouts, or aggregated outputs, PSD enables a single-model distillation setup while channeling external oracle knowledge through the prompt. PSD uses a single small model in two roles: a teacher that sees a privileged prompt containing the oracle's answer as reference context, and a student that sees only the task prompt. The student learns to reproduce the teacher's output distribution, absorbing oracle-guided behavior into its own weights without accessing the oracle's internals. On LoCoMo, PSD-trained Qwen3-0.6B, 1.7B, and 4B match or exceed GPT-4.1-mini on downstream retrieval at a fraction of the deployment cost, with off-policy PSD achieving the strongest results across most conditions. We further show that this memory-construction capability transfers out-of-distribution to LongMemEval, despite the students being trained exclusively on LoCoMo with no exposure to LongMemEval data.

Agentic AI extends LLM security beyond generated content to persistent state, autonomous actions, tool use, and interactions with humans and other agents. Existing threat classifications often emphasize individual dimensions, obscuring connections among entry points, affected components, and security consequences. The known threat landscape also differs from the coverage demonstrated by empirical research. Through a structured review of 66 studies published from 2022 to 2026, we introduce T={S, B, P, A}, a cross-dimensional representation linking affected functional or system surfaces {S}, interaction or trust boundaries {B}, violated security properties {P}, and empirically examined architectures {A}. We analyze 22 artifact-backed red-teaming studies and 11 representative security benchmarks to characterize empirical coverage and evaluation maturity. Within the selected studies, evidence concentrates on prompt/reasoning, memory, and tool-mediated attacks, predominantly in single-agent settings. Persistent, Human--Agent, complex multi-agent, systemic, and long-horizon threats receive less coverage. These findings describe the selected corpus rather than establish gaps across all empirical research. Heterogeneous metrics, limited adaptive defense evaluation, architectural imbalance, and incomplete execution-state capture further constrain comparison and reproducibility. We derive 13 open research questions to guide more systematic, architecture-aware, and reproducible security evaluation of agentic AI.

Multimodal large language models (MLLMs) incur substantial computational overhead due to the reliance on hundreds of visual tokens to represent images. While token pruning has emerged as a promising approach to reduce the inference cost of MLLMs, existing methods typically reassign position embeddings to the retained tokens using either sparse or continuous position embeddings, each introducing distinct limitations. Sparse position embeddings tend to decrease the attention value allocated to visual tokens, thereby degrading the perception capability of MLLMs, whereas continuous position embeddings disrupt the original spatial correspondence of visual tokens, leading to weakened grounding capability. To mitigate this issue, we perform layer-wise analysis of the language decoder and observe that intermediate layers play a critical role for maintaining the grounding capability of MLLMs under token pruning. Based on this observation, we propose a layer-aware position embedding strategy, which switches to sparse position embeddings at grounding-sensitive layers while maintaining continuous position embeddings elsewhere. Extensive experiments across representative pruning methods and diverse benchmarks demonstrate that our approach improves the comprehensive multimodal performance of pruned MLLMs compared with standard sparse and continuous position embeddings.

FORM is a domain-specific symbolic manipulation language widely used in particle physics for processing the very large algebraic expressions arising from multi-loop Feynman diagram calculations. Despite its central role in precision theoretical physics, no artificial-intelligence tooling exists, to our knowledge, for assisting physicists in writing FORM code. We show that contemporary large language models (LLMs), including frontier models with hundreds of billions of parameters, achieve a zero-percent execution pass rate on our instruction-following and tutorial-style FORM tasks without documentation in a single attempt, establishing FORM as a genuine zero-shot language for LLMs at the time of writing. We then present a verification-driven data generation pipeline that uses the FORM binary itself as an execution oracle to produce and validate a corpus of 4,633 training examples spanning deterministic computations, open-ended programs, tutorial code, and knowledge question-answer pairs. Fine-tuning a compact open-weights model (Qwen3-8B) with quantized low-rank adaptation (QLoRA) yields a specialist that, evaluated on four complementary benchmarks (840 tasks, single attempt each), decisively outperforms frontier models with up to 756B parameters in execution rate and in strict, FORM-verified output matching on the larger benchmarks, and remains statistically indistinguishable from them on the smaller, harder ones. General reasoning and coding capabilities are preserved within 2.6 percentage points.

Natural-language service requests can require a language-model decision before execution starts, consuming part of the request's latency budget. We integrate Jev's decision-oriented application programming interface (API) into edge service orchestration to reduce this overhead while retaining service completion. The integration extracts four to eight bounded intent fields and applies a shared validator, admission policy, and scheduler, accounting for decision waiting throughout the request timeline. We compare Jev, two self-hosted decision models, and three hosted large language models (LLMs) on 8,280 verified requests and on a live admission path with modeled execution and a real optical character recognition service. Across 33 test conditions, Jev reduces median decision latency by 22.7-64.5% relative to the fastest LLM. This latency barely moves with input size, contract width, or catalog size. On four-field contracts, Jev's API fees per correct decision are 59.7-80.9% lower at a cost of a few exact-match points, while wide contracts mark the limit of the substitution. Receiving the service catalog with each request, Jev names unseen services as accurately as known ones. On the live admission path, Jev keeps 0.91-0.95 of requests exact and on time at loads where the LLMs fall below 0.1. Since caching repeated descriptions gives the interpreters nearly the same latency, Jev's gain lies in fresh decisions. These results support decision-model substitution for latency-bound admission on bounded contracts.

Reasoning language models are trained to produce solutions, not to refuse them, and this bias persists when the problem they are handed is false. Asked to prove a corrupted theorem, a strong model will typically comply and produce a confident derivation of something untrue. We present Euston, an 8B mathematical claim-verification model trained to resist exactly this. Training data were generated with GraphSynth, a probabilistic factor-graph generator that couples attribute-level diversity to decode-time structural masking and span-synchronized verification, yielding 3{,}026 matched true/corrupted statement pairs (6,052 statements) drawn from arXiv papers spanning 2010--2025. We fine-tuned DeepSeek-R1-8B with GRPO under a rule-based, zero-API reward for 189 steps on four H100 GPUs. On a balanced 200-true/200-false held-out split, balanced accuracy rises from 29.50% to 63.75% and the discrimination gap---the difference between the rate of calling false statements false and the rate of calling true statements false moves from -0.5% (z=-0.1) to +27.5% (z=+6.0). Critically, the gain is not purchased with general mathematical ability: AIME 2026 accuracy under official semantics is 65.00% against a 69.17% base, a difference of -4.17% that is not statistically significant, whereas an earlier run of the same recipe on a smaller GraphSynth corpus collapsed to 40.00%. Median response length also falls from 19,217 to 18,296 tokens and the truncation rate from 25.8% to 8.3%, so the improvement does not come from thinking longer. We report the result together with the confounds that bound its interpretation, principally the all-false composition of the official evaluation sets and the low precision implied at realistic error prevalence.

As Large Language Model (LLM) APIs become increasingly integrated into privacy-sensitive workflows, ensuring inference-time privacy without compromising task utility remains a major challenge. Existing approaches preserve most of the original semantic content to maintain downstream performance, but this also leaves exploitable cues for reconstructing the original text. This work investigates semantic decoupling, which replaces original semantics with alternative content while preserving the structure needed for LLM reasoning. Based on this idea, we propose CROSS-MAP, a bidirectional framework that maps private inputs into a different semantic domain before inference and recovers the corresponding outputs afterward. Local models are trained with multi-objective optimization to maximize semantic divergence in the mapping stage while minimizing semantic inconsistency in the recovery stage. Experiments show that CROSS-MAP reduces reconstruction success across multiple attack settings while outperforming existing baselines in utility.

Long-context inference has become a defining capability of large language models, but exact dense attention remains costly due to its quadratic scaling with sequence length. Block-sparse attention offers a hardware-friendly alternative by routing each query block to a small set of relevant key blocks, yet accurate training-free block routing remains difficult. Existing routers often pool post-RoPE token representations, which entangles semantic aggregation with RoPE-induced geometry and attenuates local positional cues through high-frequency phase cancellation. To resolve this mismatch, we propose Semantic-Geometric Decoupled Routing, a training-free block routing framework that shifts semantic aggregation to the pre-RoPE space and reconstructs geometric bias with an offline structural prior and relative block distances. This decomposition yields an explicit closed-form block routing score without token-level search or post-hoc calibration. Experiments on long-context text and video tasks show that our method approaches full-attention accuracy across 4K--128K contexts, keeps routing overhead below 3.4 ms, and achieves a 5.03$\times$ speedup over FlashAttn at a 128K context length.

Multilingual code language models transfer skills across programming languages (PLs), but whether any PL occupies a privileged pivot position remains contested: geometric analyses point to C-family languages and Go, while behavioral evidence highlights Python. We revisit this question under controls for representational anisotropy and length variation across PLs, two confounds that compromise prior cosine-based analyses. Across three code models on multilingual competitive-programming data, we study three views of cross-PL organization: pairwise PL geometry, PL-English alignment, and pivoted retrieval through candidate PL representation spaces. The results are relation-dependent. Code-code geometry reveals structured language regions but no universal center; code-English alignment favors high-level scripting languages; and pivoted retrieval favors different intermediate spaces for code-to-code and English-to-code transfer. These findings suggest that Python's special role is better understood as English-facing affinity than as universal geometric centrality.

The availability of Large Language Models (LLMs) reshaped scientific discourse at a linguistic level. LLMs are assumed to homogenize academic writing, flattening it into a single generic lexical register. To understand how CHI writing has changed since the public release of LLMs, we analyzed full texts of 14,262 archival papers across all 44 CHI proceedings from 1982 to 2026, measuring readability, register, lexical diversity, and marker words typically produced by LLMs. We find that prose did not homogenize, while vocabulary grew more varied, and sentence rhythm remained irregular. CHI prose changed more between 2016 and 2026 than in other decades toward greater density, and reading ease has declined since 2022. The word-level shift began before any author used LLMs, so LLMs did not start the change but accelerated it. Reflecting on the history of CHI papers, we discuss what may have caused changes in prose and how LLMs accelerated them.

Step-level reasoning evaluators are commonly based on autoregressive language models, whose causal attention restricts each step representation to the problem, previous steps, and the current step. Yet, when the complete solution is available, the validity of an earlier step may become clearer only through its downstream consequences. We validate this hypothesis through a controlled 54-run comparison of causal and bidirectional LLaDA evaluators at 1B--3B scale, changing only the self-attention mask, and find bidirectional attention yields consistent improvements. Building on this finding, we introduce \prm{}, an 8B bidirectional evaluator that reaches 88.8 step-level F1 on MR-MATH-invalid and 83.8 on the out-of-distribution MR-GSM8K original-question subset, outperforming ReasonEval-Llemma-34B by 11.3 and 10.3 F1 points, respectively. \prm{} also remains effective when evaluating incomplete reasoning traces in online settings, outperforming the strongest baselines on both benchmarks by a large margin. We further show that \prm{} provides an effective training-data selection signal, improving Mistral-7B performance on MATH-500.

Chain-of-thought (CoT) can sound plausible yet be unfaithful to the model's underlying reasoning. Most prior work probes CoT faithfulness through input--output behavior or input attributions, leaving internal computation largely underexplored. We instead cast faithfulness as internal concept grounding: Does a large language model's (LLM) CoT reasoning engage the same internal concepts that support the LLM's direct prediction, and do the shared concepts causally drive its answer? Encoding a prediction pass and a CoT pass with a single shared sparse autoencoder (SAE), a reliable approximator of the latent concepts LLMs use, makes their internal concepts directly comparable. We introduce three correlational metrics of concept-level alignment and a causal metric, $Δp$, which ablates the shared concepts and measures the drop in answer probability. Across five LLMs and four datasets, concept alignment is generally high, as indicated by the correlational metrics; yet these only identify which concepts are shared, not how much they causally contribute. $Δp$ fills this gap: causal faithfulness varies substantially with model depth, peaking at mid-to-late layers rather than the final ones, and model scale reshapes the layer-wise profile. Moreover, causally important shared concepts are not always verbalized in the CoT. These dissociations suggest that faithfulness cannot be reliably assessed from surface-level or representational correspondence alone; assessing it requires causal tests of whether the internal concepts underlying a CoT actually drive the model's prediction.

The distinction between the spirit and letter of the law is a central issue across research and everyday life, and a growing concern for building safe, intelligent machines. What is this distinction based on, and how can we develop machines that follow the intention behind a rule? We used targeted adaptation that made large language models prioritize the spirit or letter of the law. With minimal modifications, our method significantly changed LLM behavior across diverse measures, novel vignettes, real-world scenarios, and influential legal cases. An analysis of model internals revealed a low-dimensional space with three interpretable dimensions matching a formal pre-specified framework for the geometry of legal concepts. These findings show how legal thought in LLMs may be organized and directed.

This thesis proposes a framework for "responsible intelligence" to address AI's critical challenges in safety, ethics, and cultural sensitivity. It advances three core areas: First, it improves domain adaptation in specialized fields using active learning and graph-based knowledge to reduce hallucinations. Second, it enhances ethical rigor via a novel decoding-time alignment mechanism that proactively blocks harmful text generation in real-time. Finally, it ensures cultural and multilingual safety through language-specific steering that respects diverse linguistic and social norms. Ultimately, this work provides a blueprint for building next-generation AI that is contextually knowledgeable, ethically sound, and culturally adaptable.

Uncertainty quantification (UQ) is increasingly recognized as an important component of reliable machine learning in medical imaging, yet practical resources connecting UQ theory, implementation, and evaluation remain limited. We developed Uncertainty Quantification in Medical Imaging Analysis (UQMIA), an open-access, hands-on tutorial comprising 19 sessions covering major UQ approaches, including variational inference, Monte Carlo dropout, deep ensembles, evidential deep learning, and conformal prediction, together with methods for evaluating uncertainty reliability. The tutorial is designed for researchers and practitioners with experience in medical imaging, progressing from foundational concepts to implementation and evaluation, with notebooks executable through Kaggle. Beyond presenting the tutorial, we introduce a framework using large language models (LLMs) to evaluate whether technical educational resources contain retrievable and usable knowledge. Using 100 four-option multiple-choice questions derived from primary methodological literature, we evaluated 20 instruction-tuned LLMs from six model families with and without retrieved tutorial context. Retrieval of UQMIA improved accuracy in 18 of 20 models, increasing mean accuracy from 0.680 to 0.742 (+0.062; Holm-adjusted P=0.00032), and improved area under the receiver operating characteristic curve (AUROC) in 18 of 20 models, increasing from 0.725 to 0.788 (+0.063; Holm-adjusted P=0.00032). UQMIA provides an accessible resource for learning UQ in medical imaging and demonstrates an LLM-based approach for evaluating technical educational material. The tutorial is available at https://benyamin-gheiji.github.io/Uncertainty-Quantification-Medical-Imaging-Analysis/ and its source code and notebooks at https://github.com/benyamin-gheiji/Uncertainty-Quantification-Medical-Imaging-Analysis/.

Cosine similarity is widely used to analyze transformer representations, implicitly assuming that similarity reflects task-relevant structure. We study when this assumption fails in dialogue-conditioned large language models. Across three 7-8B chat-tuned models, ambient cosine similarity substantially underestimates linearly decodable persona structure on the same hidden states; numerically, linear probe AUC is in the 0.73-0.97 range while cosine kNN is in the 0.56-0.77 range on a 30-class task. A low-dimensional supervised subspace recovers much of this gap, whereas a matched-rank PCA subspace does not and in some cases degrades performance. This mismatch is regime-dependent: it is absent in single-sentence sentiment classification (SST-5), and a matched-cardinality control rules out attribute cardinality as a confound. The gap does not systematically increase across dialogue turns, and the task-aligned subspace remains stable over time. However, two of three models violate a pre-registered within-subspace separability invariance criterion (|Delta AUC| <= 0.03), and one model violates a pre-registered turn-invariance criterion (|Delta L| <= 0.05). These results show that cosine similarity can fail to reflect task-aligned structure in dialogue representations even when that structure is linearly accessible.

Bilingual language models (LMs) offer a controlled setting for studying how training conditions shape second-language (L2) behaviour, but prior work typically varies exposure structure, scale, and architecture at once, making it difficult to attribute effects to any single factor. We introduce Beetle, a controlled language model pretraining framework in which tokeniser, target language, training budget, and exposure structure are each independently manipulable, enabling systematic and comparable experimentation of training conditions. Using Beetle, we train and release 285 bilingual and 45 monolingual open-source LMs with rich checkpoints across a range of exposure schedules, data scales and first languages (L1s) to study multilingual pretraining and computational modelling of bilingualism and second language learning. Evaluating models on human bilingual and second language reading-time prediction and grammaticality judgement tasks, we find that staged and temporally structured curricula consistently improve alignment with language learner reading time compared to balanced bilingual training, with the largest gains at smaller data scales and for typologically closer language pairs. The Beetle models are well suited tools to help move computational psycholinguistics beyond its prevailing monolingual, English-centric focus toward models of human bilingual processing, to study cross-lingual learning dynamics, while supporting community-based development of controlled model families.

Effective LLM tutoring depends on correctly identifying the specific error in a student's reasoning before generating feedback. We study this problem in propositional-logic proof tutoring, where student actions can be checked against formal inference rules. We introduce a verifier-grounded architecture that separates diagnosis from language generation. Using 600 balanced student actions, we compare a zero-shot LLM detector, a fine-tuned detector, and a symbolic verifier. Each diagnosis is processed by shared rationale and feedback agents, isolating the effect of the initial diagnosis. The zero-shot detector achieves a macro-F1 of 0.191; fine-tuning raises this to 0.709 but retains systematic errors between structurally related classes. Rationales generally preserve the diagnosis supplied to them, showing that an incorrect diagnosis can be faithfully propagated through the pipeline. Feedback can likewise remain faithful to its rationale, non-revealing, and pedagogically appropriate while addressing the wrong error. Verifier-grounded feedback achieves the highest diagnostic correctness, and expert ratings largely uneven with the automatic feedback evaluations. These findings show that apparent feedback quality can conceal upstream diagnostic errors and that faithfulness must be evaluated separately from correctness. Our code is publicly available

Scanned mail, uploaded PDFs, and consolidated attachments often arrive as page streams that must be split into individual documents before downstream classification, extraction, or routing. Zero-shot large language models can detect document boundaries without task-specific training, but standard Page Classification (PC) and Boundary Decision (BD) formulations resolve only one boundary per model call. We introduce Multi-Split Boundary Decision (MSBD), which predicts multiple boundaries within a page window in a single call, reducing the number of inference requests. We evaluate MSBD across multiple language models, document collections, input modalities, and window sizes. The results reveal a model- and corpus-dependent operating range in which MSBD preserves strong segmentation accuracy while substantially improving inference efficiency, followed by a sharp decline at larger windows. MSBD provided the strongest overall accuracy--efficiency trade-off, while large windows expose distinct over- and under-segmentation behavior across models. These findings show that multi-boundary prediction can make zero-shot page stream segmentation more efficient when the window size is selected for the target corpus.

Process discovery rarely yields a single coherent process structure. For analysis, a common step is to cluster process variants based on structural similarity and then assign business meaning to the resulting groups. Since these partitions are not derived from the organization's goals, analysts must manually interpret and consolidate variants into business-meaningful categories. This judgment-intensive step becomes increasingly difficult as the number and complexity of variants grow. In this paper, we propose a goal-driven approach to variant categorization that reverses this workflow. We first author an organization's goal model that predefines the categorization axis. Each variant is transformed into a textual narrative describing its behavior, and a Large Language Model (LLM) interprets it in the context of the goal model and assigns the variant to the most appropriate category. LLM-based semantic reasoning connects low-level process behavior with analyst-defined business goals. We instantiate this approach end-to-end and evaluate it on three public logs differing substantially in scale and behavioral diversity. Goal-model guidance yields partitions that differ from those produced by unguided induction and respond to controlled edits to the declared alternatives, at the cost of authoring a goal model.

The growing elderly population gives rise to an urgent need for intelligent support systems, particularly for individuals with Mild Cognitive Impairment (MCI). This paper presents PV-Care, a proactive AI-driven assistance scheme that integrates wearable electroencephalogram (EEG) sensing with visual environmental perception to provide real-time, context-aware voice assistance for MCI users. Unlike traditional assistant systems that passively wait for user commands, PV-Care actively initiates helpful interactions based on the user's detected brain states, including Learning, Memory Recall, and Resting, using a novel deep neural architecture named Spatial and Frequency Refinement Network (SFR-Net). By combining EEG-based cognitive-state recognition with AI-based visual analysis, PV-Care generates structured "4W-UT" prompts to guide the output of large language models (LLMs). Simulation results and user studies validate the high accuracy of the proposed SFR-Net and the effectiveness of PV-Care's context-aware assistance. These results indicate that PV-Care is a feasible and promising solution for MCI caring.

The success of Large Audio Language Models has driven the development of massive multimodal networks exceeding billions of parameters. However, the demand for privacy-preserving, low-latency processing has shifted focus toward Small Audio Language Models (SALMs) capable of on-device execution. In this paper, we introduce Samsone, a family of SALMs designed for edge computing. Our core model, Samsone-134M, establishes a new state-of-the-art for its size class across multiple benchmarks. We further explore the scaling laws of SALMs by introducing Samsone-99M and Samsone-356M. Despite their compact footprint, the Samsone family delivers performance competitive with models orders of magnitude larger. To foster open research and reproducibility, we train Samsone on publicly available data. We release the training code, model weights, mobile-optimized checkpoints and provide an open-source Android application to demonstrate real-time on-device inference of Samsone.