Scaling Laws & Emergent Abilities - Weekly Roundup

A controlled cross-modality framework that instantiates the same task suite in a variety of modalities to test the Convergent Emergence Hypothesis and shows that paired-mapping ICL emerges across six modalities, surpasses controlled baselines, and has correlated per-task effects across five of them.

ESRL is introduced, an architecture-aware framework that explicitly explores the expert-routing space of MoE models, and preserves high-confidence experts as anchors, and restricts stochastic routing to a plausible candidate pool, thereby retaining reliable computation paths.

ActMap is introduced, a white-box representation that compresses the generation-time hidden- state trajectory into a fixed tensor of temporal-statistic channels that preserves structure across transformer depth and pooled hidden coordinates that supports abstention, routing, and selective verification from a single generation, making it a practical primitive for scalable oversight of deployed models.

Experimental results demonstrate that DRG-MAPPO achieves a state-of-the-art win rate of 87%, suggesting that the framework effectively balances relational modeling, interpretability, and optimization stability for cooperative air combat.

The Explainability Assistant is introduced, an open-source conversational XAI system that leverages the function-calling capabilities of modern Large Language Models (LLMs) to overcome limitations and achieves 94% intent-parsing accuracy, supports flexible natural language interaction, and adapts to different ML problem types without task-specific fine-tuning.

A controlled cross-modality framework that instantiates the same task suite in a variety of modalities to test the Convergent Emergence Hypothesis and shows that paired-mapping ICL emerges across six modalities, surpasses controlled baselines, and has correlated per-task effects across five of them.

ESRL is introduced, an architecture-aware framework that explicitly explores the expert-routing space of MoE models, and preserves high-confidence experts as anchors, and restricts stochastic routing to a plausible candidate pool, thereby retaining reliable computation paths.

ActMap is introduced, a white-box representation that compresses the generation-time hidden- state trajectory into a fixed tensor of temporal-statistic channels that preserves structure across transformer depth and pooled hidden coordinates that supports abstention, routing, and selective verification from a single generation, making it a practical primitive for scalable oversight of deployed models.

Experimental results demonstrate that DRG-MAPPO achieves a state-of-the-art win rate of 87%, suggesting that the framework effectively balances relational modeling, interpretability, and optimization stability for cooperative air combat.

The Explainability Assistant is introduced, an open-source conversational XAI system that leverages the function-calling capabilities of modern Large Language Models (LLMs) to overcome limitations and achieves 94% intent-parsing accuracy, supports flexible natural language interaction, and adapts to different ML problem types without task-specific fine-tuning.

Comparing full-precision and quantized forward passes, and identifying two mechanisms that characterize pretrained quantization robustness, these mechanisms explain why quantization error that passes through numerous layers can still produce only small output changes, and are verified across models and quantization settings.

Extending LILA to dynamically allocate sparsity budgets via KS-scores yields state-of-the-art generative preservation at moderate compression, while uncovering fundamental single-layer architectural bottlenecks at higher compression regimes.

SolCloudLLM aligns sky-image patches with time-series patches and fuses their corresponding representations through bidirectional multimodal fusion, yielding a unified representation that is subsequently mapped into the embedding space of an LLM.

This work proposes a two-agent framework that first identifies the firms for which each alternative data channel is likely to be informative and then predicts revenue using firm- and channel-specific context and evaluates the framework across four commercial alternative data channels.

This paper introduces Magenta, a training-free agentic pipeline that, given only a natural-language problem, produces an answer, expresses it as a Lean 4 statement, and constructs a machine-checked proof, which achieves 100% accuracy across all evaluated olympiad benchmarks.

This study investigates the use of LLMs as post-processing tools to analyze and rank evolved expressions according to their interpretability and medical plausibility, and finds that they are better suited to comparative auditing under expert oversight than to autonomous validation.

Results are reported, showing an inverse relationship between benign and unsafe capability, strong evidence that baseline safety guardrails do not lead to downstream safety guarantees in the RAG case, and model-specific support for previous findings that even benign documents can lead to unsafe generation in retrieval-enabled systems.

The first large-scale audit of the default moderation system on Bluesky is conducted, revealing a human-AI collaborative system where labels for sexual and graphic content are applied automatically in seconds, while nuanced and high stakes labels require more human oversight, taking hours or days.

INDRA is an LLM-powered research partner enabling massive multi-archival investigations, a tool whose outputs are designed to be checked rather than trusted, and whose architecture makes the conditions of knowledge production visible and auditable.

This work constructed a large-scale Arabic Speech Question-Answering (SQA) corpus comprising over 1.5 million training samples to allow instruction tuning over a broad range of core speech tasks and designed an evaluation framework featuring diverse tasks and tailored metrics.

Whether medical research is keeping pace with the systems it evaluates is asked, and Rigour and currency are in tension, and that tension reflects model selection rather than research timelines.

This work empirically diagnose the root cause across six per-layer methods and three speech-LLM architectures and proposes a two-pool allocation that normalises encoder and LLM parameters into independent pools, and shows that joint $\ell_2$ sensitivity and the original $(\varepsilon,\delta)$-DP guarantee are unchanged.

This work analyzes over 790,000 translation outputs from 12 LLMs across 22 language pairs (LPs) and identifies 12 recurring noise patterns, which are group into formatting and content noise, and constructs TransClean, a controlled benchmark of 9,900 pairs of noisy and clean translation outputs.

Distribution-aware Language Neuron selection is proposed, which leverages pairwise relationships between per-language activation distributions over the full activation range, including negative values, to quantify each neuron's language specificity by clustering languages using pairwise overlap coefficients between their activation distributions.

By distilling a lightweight model Veilmind-4B to drive a dynamic extraction-sanitization-restoration pipeline, this approach reaches a low-leakage privacy point while preserving substantially higher response utility than existing privacy-oriented baselines, advancing the privacy-utility trade-off toward the Pareto frontier.

The way students engage with ChatGPT may be more informative than whether they use it, and this report suggests that the way students engage with ChatGPT may be more informative than whether they use it, to guide students toward effective LLM use in secure software engineering education.

LLMVul, a vulnerability-labeled dataset of LLM-generated C/C++ functions mined from real production repositories, enables reproducible research on vulnerability detection, security evaluation of LLM-generated code, and analysis of vulnerability patterns in AI-assisted software development.

Empirical results show that CamPilot outperforms state-of-the-art text-to-movie generation methods on cinematographic control and quality, highlighting the impact of professional camera design on movie generation.

This paper proposes OmniKVQuant, a training-free framework that enables 2-bit KV caches while substantially preserving performance across seven audio-visual benchmarks and provides a fused Triton decode kernel that unpacks the 2-bit cache during attention, so no dense FP16 cache is ever built.

LASP is introduced, a Language-Augmented Semantic Priors framework that leverages large language models to infer structured, solver-usable B-spline priors from procedural modeling histories, establishing a new paradigm of language-guided geometric optimization in modern CAD systems.

BlueSTAR is presented, a tiered agentic architecture for autonomous cyber defense in enterprise IT/OT networks that retains the fast containment of deterministic response for known threats while successfully defending against attacks requiring contextual and cross-cycle reasoning.

CausalArena is introduced, a unified and evolvable benchmark for causal discovery under a common protocol, and substantial ranking shifts across SCM families and protocols are revealed, showing that strong performance in one benchmark regime does not reliably transfer to others.

This work presents a training-free, alignment-free framework for corporate intelligence built on deterministic sparse seed vectors, showing how material corporate events emerge as distinct, interpretable semantic profiles, each traceable to the exact source sentences that produced it, with no domain-specific training and no LLM inference.

The model incorporates three core components: spatiotemporal embedding module, spatiotemporal fusion module, and LLM backbone, which adopts a differentiated parameter adaptation strategy to balance training efficiency and traffic data adaptability.

This work proposes Test-Time Adaptation via Stability-Aware Confidence Optimization (TASCO), a framework that incorporates local stability into confidence-based test-time adaptation while keeping the LLM frozen and improves reasoning accuracy and token efficiency across diverse LLMs and reasoning benchmarks.

The proposed greedy algorithm with alternating updates guarantees the four-peak distribution required for stable 2-bit clustering of each factor and admits closed-form initialization of cluster centers, removing the multi-restart k-means bottleneck of prior work.

A controlled inversion test over three established textual realizations of framing: evaluative lexis, agency realization, and information salience is introduced, revealing a clear separation between factual fidelity, framing recognition, and framing inversion.

NovGauge is presented, a human-anchored benchmark for fine-grained novelty assessment diagnosis, and a cascading diagnostic pipeline that verifies per-dimension correctness, evidence grounding, and logical support is proposed.

An information-weighted cross-entropy loss that rescales token-level contributions using TF-IDF statistics, emphasizing semantically informative tokens while down-weighting ubiquitous ones is presented, offering a lightweight and principled way to mitigate memorization without disrupting standard training dynamics.

This paper proposes SWRouter, a Similarity-Contractive Window Router for multi-turn large language model routing that combines a similarity-based context segmentation mechanism for prompt construction with a dual-metric evaluation framework that decouples construction accuracy from router performance.

A Database Normalization Benchmark (DNBENCH), comprising 3,275 samples for evaluating LLM-driven database normalization from 1NF to BCNF, and proposes Multi-Agent Reasoning for Schemas (MARS), which separates evidence extraction, violation diagnosis, and decomposition planning from schema generation and verification.

It is concluded that anonymization is not a one-size-fits-all solution and must be co-designed with the model and task in mind to balance privacy and utility effectively and provide a crucial baseline for developing more robust, privacy-aware AI systems.

This paper presents SurgicalRoomAgent, a voice-interactive multi-agent system for smart operating rooms based on large language models (LLMs), which achieves natural language understanding, device control, intraoperative recording, and surgical report generation through a layered architecture comprising a voice interaction pipeline and an agent core.

Results show that scaling consistently improves performance, while multitask learning acts as a beneficial regularizer primarily for smaller-capacity models.

These findings position stage-aware human-AI collaboration as a practical approach for producing inspectable, revisable, and verifiable analyses while retaining domain-expert involvement in consequential decisions.

SpecGuard is introduced, an inference-time backdoor detector that repurposes speculative decoding at zero added model-computation cost and doubles as a free, always-on signal for detecting backdoored LLM behavior.

Whether large language models (LLMs) are narrowing the variety of methods archaeologists use is evaluated, with results consistent with LLMs pushing methodological choice towards convergence, although this study cannot establish a causal effect.

EHR (Emotion Hallucination Rate), an evaluator that quantifies emotion hallucinations across six facets, and HMER (Hallucination-aware Memory-guided Emotion Reasoning), a training-free framework for emotion hallucination mitigation, which enables fine-grained mitigation across diverse hallucination facets.

This work analyzes SciLitBench, a corpus of 888 review-automation papers with 14,726 annotations, to characterize changes in methods, review-stage use, evaluation and reported limitations, and introduces PRISMA-LLM, an empirically grounded framework separating implementation disclosure from consequence-sensitive evaluation and limitation reporting.

GuardedAct is proposed, a sandbox-first remediation framework that interposes a blast-radius-aware verification layer between the LLM action generator and the production environment and achieves an overall recovery rate and a modest sandbox-induced increase in mean time to recovery.

Benchmark Radar is presented, a living database and search engine for retrieval and discovery of AI benchmarks, covering LLM evaluation, agentic and tool-use benchmarks, coding, reasoning, safety, and domain-specific evaluations.

TimelyRAG is proposed, a retriever-agnostic framework that incorporates temporal distance into ranking to align queries with version-appropriate documents, and TimelyQABench is introduced, the first benchmark for regulation-heavy domains with overlapping-evolving challenges.

A unified post-training framework that equips pretrained text-to-speech models with natural-language control over segment-level emotion and duration is proposed, highlighting post-training as a practical approach to extending existing speech synthesis models.

All three AI chatbots exhibited a high level of ability in the understanding of radiological cases with abnormal anatomy, indicating their potential use as a supplementary tool in radiological assessment and radiological diagnostic support.

This survey covers inference-efficiency mechanisms for visual and audiovisual VideoLLMs that report concrete reductions in parameter count, FLOPs per input, latency, memory, or visual and audio token count, and organize methods by the pipeline stage at which they act.

Adding an executor to the Program-Solve interface helps some open-weight models more than others even under matched formula, variable and note access, and is not a substitute for verified formulas or reliable variable extraction either way.

Across widely adopted grounding benchmarks and navigation benchmarks spanning multiple platforms, LLaDA-UI substantially outperforms Qwen2.5-VL-7B and surpasses Qwen3-VL-8B on four of six reported GUI benchmarks, establishing block-wise diffusion as a practical generative paradigm for multimodal GUI agents.

Reacting to sudden physical hazards is both a meaningful test of embodied intelligence and a hard requirement for deploying multimodal large language models (MLLMs) as the decision core of household robots, and ReactHuman is the first physics-grounded benchmark for human-like reactive decision-making.

This paper investigates the development path of large language model preference alignment techniques, focusing on the training mechanism of Reinforcement Learning from Human Feedback and its three-stage process, including supervised fine-tuning, reward model training, and PPO-based policy optimization.

Optimal learning rate and batch size in MoEs depend directly on the expert activation ratio rather than active or total parameter counts, unlocking predictable hyperparameter transfer even at 1/64 sparsity.

A checkpoint's static benchmark score cannot predict how it will respond to further training, but four tiny probe interventions can forecast downstream fine-tuning trajectories across entirely unseen model families with up to 78% lower error.

HPC-Ops Top-K is presented, a sample-guided exact selector for ragged sparse-attention score rows that outperforms the fastest verified external exact baseline and outperforms the fastest verified external exact baseline on indexer scores from Hy4-Preview.

HDA-MoE is presented, a framework that optimizes MoE execution on NMP architectures through hybrid parallel deployment and runtime scheduling and integrates an offline hybrid parallel mapping algorithm with an online dynamic and adaptive scheduling mechanism to reduce communication overhead while improving computation utilization.

StochBench is introduced, a Lean 4 benchmark of 450 graduate stochastic-processes problems at varying abstraction levels, each paired with its natural-language source that better represents domain-specific applied mathematics while remaining challenging for advanced provers.

Deep Persona is introduced, a psychologically grounded, three-layered architecture that organizes personas into hierarchical levels of observable expression, latent beliefs, and core motivational drives, for constructing highly convincing role-playing agents.

A single, architecture-independent mechanism behind the conditions where diff-based generation does win is identified: it is competitive on short, spatially localized edits, and its category-level wins concentrate in exactly the two task categories - refactoring and error-handling/edge-case fixes - with the lowest mean edit-step count in the authors' dataset.