LLM 理论进展

Diffusion Language Models (DLMs) have recently become increasingly competitive with autoregressive (AR) models, and even outperform them on certain tasks. Unlike AR models, DLMs produce output through iterative denoising without a left-to-right order. To further improve the performance of DLMs, we introduce PDC (Prefix-Denoising Consistency), a test-time self-verification method for DLMs. PDC exploits a distinctive test-time signal in DLMs under prefix conditioned regeneration, correct trajectories are more stable and reproducible than incorrect ones. Concretely, given an initially generated sample, PDC splits the sentence at an intermediate position and regenerates the remaining tokens conditioned on the fixed prefix. Across mathematical reasoning and commonsense reasoning benchmarks, PDC consistently improves upon the initial sample, outperforms independent generations under a computational constrained comparison, and is robust to different unmasking strategies and parameter settings. These results highlight prefix-conditioned regeneration as an effective DLM-specific primitive for test-time verification.

Linguistic meaning is grounded in conceptual content, from which reference to particular entities emerges as words enter discourse. To examine the processing dynamics associated with these two dimensions of meaning, we selectively disrupted conceptual or referential information in short narratives and traced the resulting effects in human self-paced reading and in the predictive and representational processing of large language models. In human reading, conceptual disruptions produced a strong but localized processing cost, emerging immediately after the distorted word, reaching an early maximum, and then declining rapidly. Referential disruptions produced weaker effects, which decreased more gradually across subsequent words, and were more strongly modulated by sentence boundaries. In the language model, both disruptions emerged immediately at the manipulated word. Contextual model surprisal showed a pattern closely paralleling human reading: conceptual disruption produced a larger, more locally concentrated effect that decayed rapidly, whereas referential disruption produced a smaller and more gradual downstream effect. Output-layer representations showed a different pattern: referential disruption produced a larger initial displacement, while both distortions were subsequently characterized by power-law decay. Together, these results provide convergent evidence for distinguishable processing dynamics of two types of meaning: conceptual information imposes a more locally concentrated integration cost, whereas referential information engages a more distributed process of maintaining discourse-level identity.

Mixture-of-Experts (MoE) architectures enable scalable and efficient large language models (LLMs) by selectively activating expert sub-networks through a routing mechanism. However, this adaptive design introduces a new attack surface: specific experts become disproportionately correlated with certain tokens (e.g., end-of-sequence), allowing adversaries to manipulate model behavior via lightweight perturbations. In this work, we present Groundhog Bit-Flip Attack (GBFA), the first bit-flip-based Denial-of-Wallet availability attack against MoE-based LLMs. By identifying and flipping routing-layer bits associated with related expert activations, we demonstrate that GBFA substantially extends the decoding token usage across three different LLM modes: conversational, reasoning, and agentic tasks, while largely preserving semantic fidelity. Across four main real-world MoE-based LLMs, manually deactivating on average fewer than 4 experts drives average output inflation to $\mathbf{5912%}$, with the majority of test samples reaching max tokens. These results reveal a robustness vulnerability of MoE architectures to bit flip, and highlight the potential of GBFA as an availability attack against LLMs.

Large language models (LLMs) excel at programming, but not when the task fixes the target language: prompted with a grammar rare in their training data, their programs usually break the grammar or fail the given specification. Enumerative synthesizers search the space of syntactically correct programs systematically guided by LLMs; the state of the art guides them by approximating LLM proposals into rule frequencies, which loses where each construct belongs and prunes every rule the proposals miss, exactly when the proposals are wrong. We present Narcissus, a synthesizer that keeps the proposals as syntax trees and scores each expansion of a candidate program in its context: does a proposal with the same surrounding structure continue the same way, and does the expansion rebuild a fragment the proposals repeat? A regularization term keeps every rule reachable, so wrong proposals delay the solution but cannot hide it. Across five domains and two search backends, Narcissus beats static guidance at every budget and consistently outperforms re-prompting the LLM to fix its own proposals; it reaches proposal-like programs an order of magnitude sooner and solves $40%$ of ARC tasks where the raw proposals solve $13%$, all without a single LLM call during search.

Large language models (LLMs) are increasingly used as evaluators to assess output quality and preference alignment, yet providing reliable guarantees of agreement with human judgments remains challenging. Recent work introduces confidence-thresholding methods that provide such guarantees for pairwise comparisons, relying on the assumption that higher estimated confidence implies lower disagreement risk with humans. However, this assumption can break down when the number of candidate responses increases, since distributing probability mass across many alternatives can distort confidence estimates. To address this issue, we propose a Localize-Then-Decide framework. First, conformal prediction localizes a small shortlist that contains the human-preferred response with high probability. Then, a calibrated confidence-based rule selectively chooses a single response from this shortlist or abstains. This design restores the monotonic relationship between confidence and disagreement risk and enables high-probability agreement guarantees. Experiments with multiple candidate sizes across several datasets and judge LLMs demonstrate that our framework consistently achieves higher guarantee success rates and substantially higher coverage than single-stage baselines.

The rapid advancement of Large Language Models (LLMs) makes it increasingly difficult to distinguish human writing from machine-generated text. Training-free detection offers a scalable solution, yet common confidence-based metrics mainly measure average token probabilities and often miss the signal fluctuations that characterize human writing, which we call "generative vitality". Spectral analysis offers a way to capture this vitality, but its mechanism and practical boundaries remain underexplored. In this paper, we analyze spectral detection from both theoretical and empirical perspectives. We connect spectral energy to variance in proxy log-probability trajectories and explain how broader human token choices create the fluctuations used by frequency-domain indicators. We further show that the strength of this signal depends on text length and sampling range: spectral evidence is clearest for long, continuous, constrained generation, while short, fragmented, mixed, and edited settings require complementary confidence and fluctuation views. These findings clarify when frequency-domain detection works and provide guidance for future multi-dimensional detector design.

Large language models increasingly mediate access to information, and a growing body of work asks whether they cite the sources behind their outputs. That work treats citation as a verification device and applies it to textual documents. Scholarly citation serves two further functions, credit and provenance, and it applies to data as much as to text. This paper argues that data citation for large language models is an open challenge, distinct from document-level citation grounding and harder to solve. We ask how such models should cite data so that outputs stay verifiable, provenance stays traceable, and credit reaches data creators and curators. We set out three research directions. Training data attribution has to turn influence estimates into references for corpora absorbed into model parameters. Data citation at inference time has to identify datasets, subsets, and query results at the right granularity and fixity. Citing knowledge graph facts has to define what a reference to a single triple denotes and how credit propagates along provenance. Progress on all three depends on joint work across the database, information retrieval, knowledge representation, and artificial intelligence communities.

Automatic Speech Recognition (ASR) is typically evaluated using Word Error Rate (WER), which poorly reflects semantic similarity. While embedding-based metrics correlate better with human judgments, the respective roles of encoder and decoder-based Large Language Models (LLMs) remain underexplored. This paper presents a comparative study of both families for ASR evaluation. We analyze BERTScore and SemDist across different LLMs, layers, and pooling strategies, showing that both metrics can achieve strong correlation with human judgments when properly configured. For decoder models, we investigate generative LLMs in two settings: pairwise hypothesis selection via prompting and direct qualitative error classification. Our results show that encoder-based metrics remain highly competitive, while generative LLMs perform strongly in hypothesis comparison and improve the interpretability of ASR evaluation.

Reasoning-oriented large language models often achieve strong problem-solving performance by generating long chains of thought, but this behavior substantially increases inference cost and latency. In contrast, instruction-tuned models tend to answer more concisely, yet often lack comparable reasoning ability. This accuracy-efficiency mismatch motivates a lightweight approach that combines the strengths of both models without full model retraining. In this paper, we propose GRIP (Granular Reward-guided Interpolation of Parameters), a reward-guided parameter interpolation framework for efficient reasoning. Given a reasoning model and an instruction model with identical architectures, GRIP assigns learnable interpolation ratios to individual modules and optimizes only these ratios while keeping both source models frozen. The interpolation ratios are trained with a reward signal that favors responses that are both correct and concise. Experiments show that GRIP achieves a better accuracy-efficiency trade-off than fixed or search-based merging baselines and further reveals module-wise fusion patterns associated with efficient reasoning.

Large language models (LLMs) increasingly assess generated content, giving rise to the LLM-as-a-Judge paradigm. These systems now score outputs, filter content, and gate iterative refinement in production pipelines, where each judgment is often assumed to be independent of earlier evaluations. We test this assumption using three prompt conditions: no metadata, revision framing, and anchored metadata containing revision, attempt, and prior-score fields. We show that prior scores, even when included only as context metadata, anchor judgments and systematically shift ratings toward their values. Across 192,000 attempted evaluations (185,271 successful), seven out of the eight evaluated models have 95% task-stratified bootstrap intervals below zero for the total anchored-metadata effect on 20 fixed texts. Cohen's $d$, a standardized measure of the difference between score distributions, reaches an absolute value of 0.71. Token-level analysis of selected model-task probes suggests a threshold-like response pattern: introducing anchored metadata produces a marked redistribution of output-score probabilities, while changing the anchor value within the tested below-threshold range produces comparatively little additional variation. On categorical industry data with human-labeled ground truth, anchored metadata blocks 48% of error corrections and flips 10.18% of correct judgments toward an assigned wrong label, demonstrating the bias extends beyond numerical scoring to categorical decisions. Neither Chain-of-Thought nor a metadata-disregard warning reduces the total effect, although the warning improves the paired accuracy effect relative to baseline in the industry experiment. Reliable LLM evaluation demands careful context engineering rather than an assumption of impartiality. Effective mitigation must be validated for the intended model and task or domain.

Correcting flawed reasoning traces remains a significant challenge for Large Language Models (LLMs), whose autoregressive generation can propagate early mistakes into subsequent reasoning. We introduce DCGC, a Masked Diffusion Model (MDM) framework for global correction that uses an imperfect solution draft from an upstream solver as auxiliary context. DCGC combines task-specific Supervised Fine-Tuning (SFT) with a novel inference-time mechanism called Dynamic Dual-CFG. This mechanism separates problem-only and joint problem-draft branches and scales the draft-conditioned residual using a relative confidence gap. Across math, code, and knowledge reasoning benchmarks, DCGC outperforms standard sampling and simpler CFG variants, with additional results suggesting transfer to different diffusion backbones. In test-time setting where ground-truth failure labels are unavailable, DCGC improves full test set accuracy by correcting low-consensus upstream outputs, highlighting its utility as a verifier-free global correction module for difficult reasoning instances.

Automated novelty judgment can accelerate scientific discovery by enabling efficient evaluation, refinement, and comparison of research ideas. While large language models are increasingly adopted for this task, we investigate a previously overlooked limitation in their judgment capabilities: despite generating reasoning rationales that closely mirror those of human experts, their final novelty judgments often diverge substantially. We demonstrate that this miscalibration stems from a systematic bias towards judging ideas as "medium novel". To mitigate this, we propose Think-Probe-Respond (TPR), a lightweight approach that probes latent novelty judgments from hidden states during the reasoning phase and uses the probed judgments to condition the final response. Across strong baselines, TPR improves novelty judgment performance by 22.30% and successfully mitigates the prevalent "medium novelty" bias.

Crosslingual evaluation of language models that enables fair comparisons remains a fundamental challenge in multilingual NLP. Existing studies adopt a variety of downstream tasks and intrinsic metrics with different theoretical justifications, yet there has been little empirical investigation into whether these approaches yield meaningful crosslingual conclusions. We systematically examine crosslingual evaluation approaches using controlled monolingual language models trained on parallel data with varying tokenizer vocabulary sizes and model sizes, and further validate our findings on multilingual LLMs. We further discuss challenges in achieving comparable downstream evaluation across languages. Our results show that several widely used normalized metrics introduce crosslinguistic biases rooted in tokenization, encoding, and orthographic differences. In contrast, sentence-level negative log-likelihood computed over semantically equivalent sequences provides more meaningful and consistent crosslingual comparisons.

Vision-Language-Action (VLA) models encode visual observations as flat 2D patch tokens that carry no intrinsic geometric structure, and augmenting them with dense monocular depth injects per-pixel scalar values that encode neither surface orientation nor geometric confidence. This leaves the policy with limited structured spatial reasoning for action prediction. We propose GaussVLA, a Mamba-based VLA that incorporates two custom modules: Gaussian Spatial Tokenizer (GST) to lift frozen semantic and depth features into compact 3D Gaussian tokens, pools geometrically salient regions with learned queries, and Depth-Aware Chain-of-Thought (DA-CoT) that performs structured, non-autoregressive geometric reasoning under language and flow-time conditioning. Across both simulation and real-world evaluations, GaussVLA demonstrates strong spatial-manipulation performance while remaining parameter-efficient. On LIBERO, it achieves 93.5% average success and 100.0% success on the Spatial suite with only 200M parameters, improving over SpatialVLA by 19.7% relative average success while remaining significantly more parameter-efficient.

In multi-view anomaly detection, more cross-view information can actually hurt. When multiple inspection views are naively fused in a reconstruction-based pipeline, normal cues from intact views propagate to the decoder, which faithfully reconstructs anomalous regions, collapsing the reconstruction gap the detector depends on. We call this failure mode cross-view information leakage and show that effective multi-view fusion must explicitly restrict the information reaching the decoder. Building on this insight, we present GLAD(Global-Local Attention Driven framework), the first framework combining vision foundation model features with local and global cross-view fusion for multi-view anomaly detection. The Multi-view Merging Attention (MMA) module performs local cross-view fusion at linear complexity with learnable view importance weighting and token-wise gating, letting each view selectively incorporate fine-grained evidence from other views at $\mathcal{O}(N)$ cost. The Object-Guided Attention (OGA) module captures global context by aggregating class tokens from all views into a single object-level representation and broadcasting it back to patch tokens via temperature-scaled sigmoid gating, replacing the original patch representations rather than adding a residual to preserve the reconstruction gap. Experiments on Real-IAD and MANTA-Tiny show that GLAD outperforms state-of-the-art methods across sample-, image-, and pixel-level metrics, confirming that principled information restriction is key to multi-view anomaly reasoning.

Systematic reviews, scoping reviews, mapping studies, and related evidence syntheses are increasingly difficult to conduct with fully manual workflows as search volumes, update cycles, and synthesis requirements continue to expand. At the same time, artificial intelligence, machine learning, and large language models are rapidly entering review practice across query formulation, screening, extraction, categorization, appraisal support, and reporting. Yet the empirical evidence remains uneven, task-dependent, and insufficient to justify unconstrained automation. Existing standards such as PRISMA 2020, PRISMA-S, PRISMA-ScR, PRISMA-P, PRESS, and SWiM remain essential, but none provides an end-to-end operational standard for when AI use is methodologically appropriate, how it should be validated, which review decisions must remain human-led, and how AI involvement should be reported so that readers can audit it. This paper proposes ARISMA, an AI Reporting and Integration standard for Systematic Methods and Analysis. ARISMA treats AI as an inspected, benchmarked, logged, and reversible assistant rather than an autonomous reviewer. It is built around one governing principle: every consequential scientific decision must remain human-interpretable, human-auditable, and human-accountable. The paper contributes a lifecycle taxonomy, process guidance, stepwise recommendations across the review pipeline, a governance and provenance model, a tool-support framework, an AI-integrated reporting checklist, and a validation matrix. It also addresses legal, privacy, infrastructure, and sustainability considerations. The framework was iteratively refined through structured expert consultation. The result is a practical and auditable guideline for responsible AI-assisted evidence synthesis.

The growing adoption of large language models (LLMs) has raised increasing concerns about the energy consumption and environmental impact of inference. This paper presents a systematic empirical study of decode-phase energy consumption across representative open-source LLMs employing Multi-Head Attention (MHA), Grouped Query Attention (GQA), and Grouped Query Attention with Sliding Window Attention (SWA) to characterize how attention architecture influences decode-phase energy consumption under varying inference workloads. We evaluate four models across different context lengths, batch sizes, and generation workloads while measuring GPU energy using NVIDIA hardware counters. We examine the effects of context length, attention mechanism, Key-Value (KV) cache growth, and batching on decode-phase energy consumption. Results show that attention mechanism is the primary factor governing how decode energy scales with context length. MHA models exhibit substantially steeper energy growth than GQA models, whereas GQA with SWA maintains nearly constant energy consumption. We further show that model size primarily determines absolute energy consumption, while batching reduces both energy per generated token and request latency by up to 87%. These findings provide practical guidance for selecting energy-efficient LLM architectures and inference configurations.

Although Speech Large Language Models (SpeechLLMs) excel at speech understanding and generation, their capacity for fine-grained, temporally aligned outputs remains underexplored. Our work addresses this gap by enabling SpeechLLMs to jointly model speech content and temporal structure, effectively transforming them from ``content understanding machines" into ``temporal-aware content understanding machines". Specifically, we replace traditional absolute timestamps with relative timestamps, achieving a more compact vocabulary and stronger generalization capabilities. To efficiently infuse timestamp prediction ability into pre-trained large language models, we introduce a hybrid fine-tuning strategy: full-parameter fine-tuning of the timestamp-augmented embedding layer and language model head, combined with LoRA fine-tuning of the decoder layers. Moreover, we design a masked timestamp training objective, preventing the model from over-relying on ground-truth timestamps, and thereby enhancing robustness against noisy real-world annotations. Extensive experiments demonstrate that our approach achieves significant improvements in timestamp prediction accuracy while maintaining strong speech transcription performance.

Large language models (LLMs) are increasingly deployed as AI analysts to process financial disclosures and support AI-assisted investment decisions. Yet such systems are usually evaluated by what they can retrieve, not whether retrieved information affects their judgments. We identify a retrieval-integration gap in long-context financial analysis. Holding focal-firm information fixed and varying only unrelated context from 2,000 to 128,000 tokens, we find that a risk disclosure's influence on investment judgments falls to the experimental noise floor even as direct retrieval remains accurate. The pattern replicates across model families and judgment tasks and in experiments removing real disclosures from actual 10-K filings. More capable models postpone but do not eliminate the gap. Causal memory interventions show that compressed summaries and source-text lookup jointly transmit disclosures into judgments. Workflow architecture determines whether this transmission succeeds: chunk-and-summarize pipelines evict relevant information, whereas a targeted, structured restatement adjacent to the decision restores its influence. AI analyst performance is therefore jointly determined by model capability and workflow architecture. Retrieval-based evaluations can certify systems whose investment judgments ignore information they demonstrably retrieved.

An intelligent system does not merely reason: it governs its own reasoning - how much to compute, when to stop, which module to activate. Can that role be played by a dynamic internal field - a low-dimensional homeostatic state with explicit physics and certified stability - that modulates cognition without performing it? Ours is a field on the module graph governed by a family of PDEs on the graph Laplacian, advancing with an adaptive-depth reasoner. We certify the stability of the integrator of the whole family - an integrator certificate, not a closed-loop one. New, and proved here: a discrete Schur-Cohn criterion for Verlet with velocity coupling, necessary and sufficient per latent root, with no commutation hypothesis. The answer is threefold: substance no, structure only in part, certifiability yes. The type of the field's physics is irrelevant for accuracy: wave, diffusion, gated mixtures and a 2D Navier-Stokes substrate tie. A twenty-seed preregistered deconfounding campaign bounds the structural claim: at equalized caps the second-order effect is strong in one family (+0.087 [+0.042, +0.132], t=4.0) but is not detected in the other (+0.014 [-0.013, +0.040], n.s.), so part of the original contrast was capacity, not order; and a matched-interface GRU is indistinguishable in the first and nominally exceeds the field in the second (-0.035 [-0.067, -0.002]). What distinguishes the field is not capability but that its one-step operator admits an exact runtime stability check - a difference of kind, not of existence: learned recurrences carry certificates too, sufficient and conservative ones. A kill-gate with a positive control finds no evidence for the field as evidence accumulator (Delta AUC +0.0007 [-0.0065, +0.0079] vs a 0.03 threshold). A dynamic internal field is a viable, certifiable compute governor, but not an enhancer of cognition: it modulates, it does not think.

Large Language Models (LLMs) are increasingly being embedded into all facets of society, from search to education, industrial, and financial applications. These systems' carbon and water footprints raise important sustainability concerns, particularly with adoption rates exceeding 80% among university students, despite limited insight into the environmental impacts of individual usage. Eco-feedback interfaces offer a promising approach to encourage more sustainable behaviors, yet their role in shaping LLM users' sustainability awareness and decision-making remains underexplored. We design and deploy the interface that visualizes latency-carbon trade-offs during live LLM interactions. We study its use with undergraduate computer science students (N=89, ages 18-24), enrolled in a computing ethics course, providing an empirical look at how a technically sophisticated and values-oriented user population responds to sustainability-aware AI interfaces. We found that the likelihood of choosing the eco-feedback system significantly decreased as perceived response latency increased (p < .001), while users' willingness increased when they recognized the carbon-saving impacts (p < .01). Also, students with stronger eco-mindedness demonstrated higher baseline willingness to adopt lower-carbon modes and reported increased awareness of the environmental impacts of LLM use, though this effect diminished as latency increased. These results position eco-feedback interfaces as a promising sustainability intervention and highlight their potential as an educational opportunity to promote more sustainable LLM use among university students and beyond.

A unified audio model must recognize and understand linguistic, paralinguistic, and environmental information while supporting speech synthesis and editing. A key challenge is representation: understanding favors compact features suited to long-context modeling, whereas speech generation requires reconstructible features that preserve fine-grained acoustic detail. We introduce FireRedAudio, a general-purpose audio language model with a shared 9B-parameter LLM. To the best of our knowledge, it is the first publicly disclosed unified audio-language model to provide separate continuous input representations for understanding and generation within a single trainable autoregressive LLM. Audio to be recognized or analyzed is processed by a dedicated Audio Encoder, while speech inputs for generation use a RedAE-based pathway. The LLM directly generates text or conditions a flow-matching DiT to produce continuous acoustic latents. Through progressive multitask training, FireRedAudio supports ASR and audio understanding, with the latter extending to recordings of up to one hour, as well as zero-shot TTS, Instruct TTS, and semantic and acoustic speech editing. Its structured organization of long-form audio achieves second-level timestamp accuracy. Across comprehensive evaluations, FireRedAudio achieves competitive or leading performance in audio understanding and multilingual ASR, strong content accuracy and speaker preservation in zero-shot TTS, leading instruction following in Instruct TTS, and substantial improvements over Ming-UniAudio-Edit in both semantic and acoustic speech editing. These results demonstrate the viability of decoupled continuous input representations for unifying audio understanding and continuous-latent speech generation in a model of moderate scale. Our code is available at https://github.com/FireRedTeam/FireRedAudio.

Inspired by human multi-modal perception, Audio-Visual Multi-Modal Learning (AVMML) integrates auditory and visual information to leverage complementary cross-modal cues, enabling more robust and comprehensive scene perception. Existing studies predominantly tackle each AVMML task in isolation, which stands in stark contrast to humans' unified cognitive capacity for handling versatile perception. However, naive joint training across multiple AVMML tasks often suffers from mutual interference, arising from the intricate inter-task relationships. To address this, we propose a unified framework that simultaneously accommodates versatile AVMML tasks. Specifically, benefiting from powerful representation and generalization capabilities of large language models, we design a task-disentangled Low-Rank Adaptation (LoRA) mechanism that enables dynamic integration of both task-specific and task-shared knowledge, thereby facilitating effective multi-task collaboration. The proposed task-disentangled LoRA comprises three components: a task-general low-rank matrix, task-specific modulation matrices, and cross-task collaboration experts, which respectively capture universal audio-visual knowledge, decouple task-specific pattern, and exploit inherent inter-task correlations. By unifying explicit collaboration from both model and task perspectives, our approach not only surpasses existing unified audio-visual models across multiple AVMML tasks, but also outperforms most task-specific models on certain AVMML tasks.

Inference-time decoding methods improve LLM reasoning by exploring multiple candidate trajectories, yet treat each trajectory as atomic: either retaining it whole or discarding it irreversibly. This wastes computation on partially promising candidates whose high-quality prefixes are abandoned alongside degraded suffixes. We introduce Selective Regenerative Decoding (SRD), which routes each candidate to discard, keep, or refine only the degraded portion of the suffix while preserving the useful prefix of borderline candidates, without requiring a larger target model. Under mild assumptions, SRD achieves a provable 1.28-to-1.36-fold gain in sample efficiency over rejection sampling with strictly higher expected trajectory quality, with the gain growing as the candidate pool grows. Across MATH500, GPQA Diamond, HotpotQA, and AlpacaEval with multiple generation-reward model pairs, SRD matches Best-of-N accuracy with substantially fewer generated tokens and outperforms speculative rejection in low-compute regimes. By enabling segment-level intervention rather than whole-trajectory selection, SRD opens a previously underexplored region of the accuracy-compute tradeoff for inference-time reasoning.

Large Language Models (LLMs) increasingly generate code from natural-language prompts, making prompt engineering a key mechanism for shaping the security of generated software. Structured and security-oriented prompts are widely used to encourage safer code, yet their effects extend beyond whether detected weaknesses are simply present or absent. Using 424 security-sensitive Python tasks, we generate solutions with GPT-4o and LLaMA 3.1-8B under five prompt variants that progressively add structural and security guidance, and evaluate them with Bandit and CodeQL along two axes: generation compliance and security weakness prevalence, severity, and CWE distributions. Structured prompting substantially reduces refusals (e.g., GPT-4o invalid outputs drop from 338 of 424 to 37-52), enabling large-scale analysis, but security-oriented refinements do not consistently reduce overall weakness prevalence. For GPT-4o, stronger prompts primarily redistribute risk: high-severity findings fall (20.8% to 13.6%) while low-severity findings rise (32% to 43.5%); LLaMA shows weaker, less consistent shifts. We also observe security-driven semantic drift, where stricter prompts silently remove or rewrite explicitly requested unsafe constructs. Overall, prompt structure improves compliance but is an unreliable substitute for robust security controls in LLM-assisted development.

Scientific reasoning requires language models to retrieve specialized knowledge and incorporate it reliably into multi-step computation. Conditional memory provides an explicit lookup pathway that complements dense neural representations, but its usefulness is inherently input- and computation-dependent: retrieved information may repair missing scientific associations, yet it may also introduce distracting shortcuts or interfere with reasoning that the base model can already perform correctly. In this work, we systematically investigate when, where, and to what extent conditional memory should participate in scientific reasoning. We characterize the scientific knowledge boundary and controlled interventions on memory-enabled knowledge-circuit nodes. Based on these analyses, we propose a Knowledge Boundary-Aware Router that uses task-specific input proxies available before generation to determine whether memory is activated, which layer-stage nodes receive memory signals, and how strongly these signals contribute. Experiments on biological and chemical reasoning benchmarks, covering two backbone families and six task types, show that memory effects vary substantially across inputs, tasks, and injection locations. Compared with static and activation-rate-matched random routing, our approach more consistently preserves beneficial memory contributions while suppressing memory-induced regressions, establishing selective memory allocation as an important principle for reliable scientific reasoning.

While recent advances in data synthesis aim to curate high-quality datasets, most generation pipelines still rely on heuristic prompt-based control. This black-box paradigm provides limited insight into how individual samples interact with a model's underlying learning dynamics. To bridge this gap, we propose a circuit-grounded framework that connects training-dynamics-based data valuation with mechanistic interpretability (MI). Specifically, we conceptualize data quality along three complementary utility axes, learnability, challenge, and alignment. First, we uncover specialized model-internal circuits that causally govern these utility signals. Then, moving beyond heuristic prompting toward mechanistic control, we leverage these circuits as controllable interfaces, actively steering generation to produce utility-targeted data. Building on this capability, we introduce SAMS (Stage-Aware Mechanistic Scheduling), which schedules circuit-steered data according to the model's evolving optimization needs. Experiments on multiple-choice QA tasks demonstrate that our approach yields precisely controlled data with greater diversity than prompt-based baselines, consistently improving downstream performance and calibration. Ultimately, this work establishes a principled white-box paradigm for interpretable data generation, pioneering the use of MI not just as an analytical tool, but as a practical, controllable interface.

Urdu, the world's tenth most spoken language with 246 million speakers, remains almost entirely absent from mainstream LLM safety evaluation and nine years of WOAH proceedings. To investigate whether this absence has measurable consequences for content moderation reliability, five large language models, GPT-4o, Claude Sonnet 4.5, Gemini 2.5 Flash, Qwen-2.5, and Llama-3.1, were tested across six datasets spanning Nastaliq Urdu, Roman Urdu, English, and code-switched Urdu-English. Across the five Urdu-script datasets, label instability between original-script and English-translation classification ranged from 15.9% (Gemini 2.5 Flash) to 31.6% (Qwen-2.5), with a 'Missed-in-Urdu' rate, content flagged as harmful in English translation but passed as normal in the original script, ranging from 2.4% to 9.9% (median 4.3%). A complete enumeration of all 205 papers across nine ALW/WOAH editions via the ACL Anthology API confirms zero dedicated Urdu papers across the entire period. Results indicate that current LLMs provide uneven safety assurance across Urdu's script varieties, with smaller open-weight models showing substantially higher instability and missed-harm rates than frontier closed models.

The rapid expansion of large-scale assessments and the growing adoption of automatic item generation have intensified concerns about incidental content redundancy, where construct-irrelevant elements such as wording or contextual framing become unintentionally repetitive across items. Traditional similarity metrics like BLEU or cosine similarity, often fail to capture the nuanced structural and semantic layers that drive perceived redundancy simultaneously. This study proposes a dual-dimensional framework for Automated Item Similarity Analysis (AISA) powered by Large Language Models (LLMs), operationalizing similarity through Structured Decomposition and Semantic Relatedness. Psychometric validation indicates that LLM-derived metrics align more closely with indicators of construct-irrelevant local dependence and yield more coherent item parameter groupings than traditional text-based measures. The framework is further evaluated through its application in Computerized Adaptive Testing (CAT). Simulations reveal that incorporating LLM-based similarity constraints into item selection improves estimation stability and reduces bias with minimal efficiency trade-offs, outperforming constraints based on conventional metrics. These findings highlight the potential of LLM-powered AISA to support scalable bank curation, content-aware test assembly, and experience-sensitive adaptive testing across diverse assessment contexts.

The rapid scaling of large language models (LLMs), particularly mixture-of-experts (MoE) architectures, has intensified interconnect demands because expert-parallel execution is communication-intensive. Wafer-scale optical interconnects based on dense wavelength-division multiplexing (DWDM) offer a promising path to higher bandwidth; however, conventional microring-resonator (MRR)-based links rely on thermo-optic tuning and are therefore vulnerable to workload-induced thermal fluctuations. In this work, we present a cross-layer analysis of wafer-scale optical interconnects for MoE workloads that combines workload profiling, packet-level network simulation, and transient thermal analysis. We implement a wafer-scale topology in the ht-sim simulator and construct an Ansys thermal model of a 3D-integrated GPU/EIC/PIC stack. Our results show that transient temperature variations can exceed the tracking capability of conventional thermo-optic control loops and thereby introduce repeated tuning stalls during communication phases. The stall durations injected into the network simulation are derived directly from the thermal model rather than assumed. We further evaluate a ferroelectric-based electro-optic tuning mechanism that removes the continuous thermal-tuning requirement. In a four-layer proxy simulation across three MoE models, eliminating the tuning stalls yields speedups of 2.7x for Mixtral 8x7B, 3.8x for Qwen-MoE 14.3B, and 3.3x for LLaMA-MoE 6.7B relative to the thermo-optic case. These results indicate that minimizing photonic tuning latency is important for realizing the performance potential of optical interconnects in large-scale AI systems.