Many-Tier Instruction Hierarchy in LLM Agents

Large language model agents receive instructions from many sources-system messages, user prompts, tool outputs, other agents, and more-each carrying different levels of trust and authority. When these instructions conflict, agents must reliably follow the highest-privilege instruction to remain safe and effective. The dominant paradigm, instruction hierarchy (IH), assumes a fixed, small set of privilege levels (typically fewer than five) defined by rigid role labels (e.g., system > user). This is inadequate for real-world agentic settings, where conflicts can arise across far more sources and contexts. In this work, we propose Many-Tier Instruction Hierarchy (ManyIH), a paradigm for resolving instruction conflicts among instructions with arbitrarily many privilege levels. We introduce ManyIH-Bench, the first benchmark for ManyIH. ManyIH-Bench requires models to navigate up to 12 levels of conflicting instructions with varying privileges, comprising 853 agentic tasks (427 coding and 426 instruction-following). ManyIH-Bench composes constraints developed by LLMs and verified by humans to create realistic and difficult test cases spanning 46 real-world agents. Our experiments show that even the current frontier models perform poorly (~40% accuracy) when instruction conflict scales. This work underscores the urgent need for methods that explicitly target fine-grained, scalable instruction conflict resolution in agentic settings.

Steered LLM Activations are Non-Surjective

Activation steering is a popular white-box control technique that modifies model activations to elicit an abstract change in output behavior. It has also become a standard tool in interpretability (e.g., probing truthfulness, or translating activations into human-readable explanations and safety research (e.g., studying jailbreakability). However, it is unclear whether steered activation states are realizable by any textual prompt. In this work, we cast this question as a surjectivity problem: for a fixed model, does every steered activation admit a pre-image under the model's natural forward pass? Under practical assumptions, we prove that activation steering pushes the residual stream off the manifold of states reachable from discrete prompts. Almost surely, no prompt can reproduce the same internal behavior induced by steering. We also illustrate this finding empirically across three widely used LLMs. Our results establish a formal separation between white-box steerability and black-box prompting. We therefore caution against interpreting the ease and success of activation steering as evidence of prompt-based interpretability or vulnerability, and argue for evaluation protocols that explicitly decouple white-box and black-box interventions.

GENFIG1: Visual Summaries of Scholarly Work as a Challenge for Vision-Language Models

In many science papers, "Figure 1" serves as the primary visual summary of the core research idea. These figures are visually simple yet conceptually rich, often requiring significant effort and iteration by human authors to get right, highlighting the difficulty of science visual communication. With this intuition, we introduce GENFIG1, a benchmark for generative AI models (e.g., Vision-Language Models). GENFIG1 evaluates models for their ability to produce figures that clearly express and motivate the central idea of a paper (title, abstract, introduction, and figure caption) as input. Solving GENFIG1 requires more than producing visually appealing graphics: the task entails reasoning for text-to-image generation that couples scientific understanding with visual synthesis. Specifically, models must (i) comprehend and grasp the technical concepts of the paper, (ii) identify the most salient ones, and (iii) design a coherent and aesthetically effective graphic that conveys those concepts visually and is faithful to the input. We curate the benchmark from papers published at top deep-learning conferences, apply stringent quality control, and introduce an automatic evaluation metric that correlates well with expert human judgments. We evaluate a suite of representative models on GENFIG1 and demonstrate that the task presents significant challenges, even for the best-performing systems. We hope this benchmark serves as a foundation for future progress in multimodal AI.

Are Finer Citations Always Better? Rethinking Granularity for Attributed Generation

Citation granularity - whether to cite individual sentences, paragraphs, or documents - is a critical design choice in attributed generation. While fine-grained citations are often preferred for precise human verification, their impact on model performance remains under-explored. We analyze four model scales (8B-120B) and demonstrate that enforcing fine-grained citations degrades attribution quality by 16-276% compared to the best-performing granularity. We observe a consistent performance pattern where attribution quality peaks at intermediate granularities (paragraph-level). Our analysis suggests that fine-grained (sentence-level) citations disrupt necessary semantic dependencies for attributing evidence to answer claims, while excessively coarse citations (multi-paragraph) introduce distracting noise. Importantly, the magnitude of this performance gap varies non-monotonically with model scale: fine-grained constraints disproportionately penalize larger models, suggesting that atomic citation units disrupt the multi-sentence information synthesis at which these models excel. Strikingly, citation-optimal granularity leads to substantial gains in attribution quality while preserving or even improving answer correctness. Overall, our findings demonstrate that optimizing solely for human verification via fine-grained citation disregards model constraints, compromising both attribution faithfulness and generation reliability. Instead, effective attribution requires aligning citation granularity with the model's natural semantic scope.

CRISP: Characterizing Relative Impact of Scholarly Publications

Assessing a cited paper's impact is typically done by analyzing its citation context in isolation within the citing paper. While this focuses on the most directly relevant text, it prevents relative comparisons across all the works a paper cites. We propose CRISP, which instead jointly ranks all cited papers within a citing paper using large language models (LLMs). To mitigate LLMs' positional bias, we rank each list three times in a randomized order and aggregate the impact labels through majority voting. This joint approach leverages the full citation context, rather than evaluating citations independently, to more reliably distinguish impactful references. CRISP outperforms a prior state-of-the-art impact classifier by +9.5% accuracy and +8.3% F1 on a dataset of human-annotated citations. CRISP further gains efficiency through fewer LLM calls and performs competitively with an open-source model, enabling scalable, cost-effective citation impact analysis. We release our rankings, impact labels, and codebase to support future research.

A Very Big Video Reasoning Suite

Rapid progress in video models has largely focused on visual quality, leaving their reasoning capabilities underexplored. Video reasoning grounds intelligence in spatiotemporally consistent visual environments that go beyond what text can naturally capture, enabling intuitive reasoning over spatiotemporal structure such as continuity, interaction, and causality. However, systematically studying video reasoning and its scaling behavior is hindered by the lack of large-scale training data. To address this gap, we introduce the Very Big Video Reasoning (VBVR) Dataset, an unprecedentedly large-scale resource spanning 200 curated reasoning tasks following a principled taxonomy and over one million video clips, approximately three orders of magnitude larger than existing datasets. We further present VBVR-Bench, a verifiable evaluation framework that moves beyond model-based judging by incorporating rule-based, human-aligned scorers, enabling reproducible and interpretable diagnosis of video reasoning capabilities. Leveraging the VBVR suite, we conduct one of the first large-scale scaling studies of video reasoning and observe early signs of emergent generalization to unseen reasoning tasks. Together, VBVR lays a foundation for the next stage of research in generalizable video reasoning. The data, benchmark toolkit, and models are publicly available at https://video-reason.com/ .

Safe and Interpretable Multimodal Path Planning for Multi-Agent Cooperation

Successful cooperation among decentralized agents requires each agent to quickly adapt its plan to the behavior of other agents. In scenarios where agents cannot confidently predict one another's intentions and plans, language communication can be crucial for ensuring safety. In this work, we focus on path-level cooperation in which agents must adapt their paths to one another in order to avoid collisions or perform physical collaboration such as joint carrying. In particular, we propose a safe and interpretable multimodal path planning method, CaPE (Code as Path Editor), which generates and updates path plans for an agent based on the environment and language communication from other agents. CaPE leverages a vision-language model (VLM) to synthesize a path editing program verified by a model-based planner, grounding communication to path plan updates in a safe and interpretable way. We evaluate our approach in diverse simulated and real-world scenarios, including multi-robot and human-robot cooperation in autonomous driving, household, and joint carrying tasks. Experimental results demonstrate that CaPE can be integrated into different robotic systems as a plug-and-play module, greatly enhancing a robot's ability to align its plan to language communication from other robots or humans. We also show that the combination of the VLM-based path editing program synthesis and model-based planning safety enables robots to achieve open-ended cooperation while maintaining safety and interpretability.

Conformal Thinking: Risk Control for Reasoning on a Compute Budget

Reasoning Large Language Models (LLMs) enable test-time scaling, with dataset-level accuracy improving as the token budget increases, motivating adaptive reasoning -- spending tokens when they improve reliability and stopping early when additional computation is unlikely to help. However, setting the token budget, as well as the threshold for adaptive reasoning, is a practical challenge that entails a fundamental risk-accuracy trade-off. We re-frame the budget setting problem as risk control, limiting the error rate while minimizing compute. Our framework introduces an upper threshold that stops reasoning when the model is confident (risking incorrect output) and a novel parametric lower threshold that preemptively stops unsolvable instances (risking premature stoppage). Given a target risk and a validation set, we use distribution-free risk control to optimally specify these stopping mechanisms. For scenarios with multiple budget controlling criteria, we incorporate an efficiency loss to select the most computationally efficient exiting mechanism. Empirical results across diverse reasoning tasks and models demonstrate the effectiveness of our risk control approach, demonstrating computational efficiency gains from the lower threshold and ensemble stopping mechanisms while adhering to the user-specified risk target.

More Than Efficiency: Embedding Compression Improves Domain Adaptation in Dense Retrieval

Dense retrievers powered by pretrained embeddings are widely used for document retrieval but struggle in specialized domains due to the mismatches between the training and target domain distributions. Domain adaptation typically requires costly annotation and retraining of query-document pairs. In this work, we revisit an overlooked alternative: applying PCA to domain embeddings to derive lower-dimensional representations that preserve domain-relevant features while discarding non-discriminative components. Though traditionally used for efficiency, we demonstrate that this simple embedding compression can effectively improve retrieval performance. Evaluated across 9 retrievers and 14 MTEB datasets, PCA applied solely to query embeddings improves NDCG@10 in 75.4% of model-dataset pairs, offering a simple and lightweight method for domain adaptation.

Genomic Next-Token Predictors are In-Context Learners

In-context learning (ICL) -- the capacity of a model to infer and apply abstract patterns from examples provided within its input -- has been extensively studied in large language models trained for next-token prediction on human text. In fact, prior work often attributes this emergent behavior to distinctive statistical properties in human language. This raises a fundamental question: can ICL arise organically in other sequence domains purely through large-scale predictive training? To explore this, we turn to genomic sequences, an alternative symbolic domain rich in statistical structure. Specifically, we study the Evo2 genomic model, trained predominantly on next-nucleotide (A/T/C/G) prediction, at a scale comparable to mid-sized LLMs. We develop a controlled experimental framework comprising symbolic reasoning tasks instantiated in both linguistic and genomic forms, enabling direct comparison of ICL across genomic and linguistic models. Our results show that genomic models, like their linguistic counterparts, exhibit log-linear gains in pattern induction as the number of in-context demonstrations increases. To the best of our knowledge, this is the first evidence of organically emergent ICL in genomic sequences, supporting the hypothesis that ICL arises as a consequence of large-scale predictive modeling over rich data. These findings extend emergent meta-learning beyond language, pointing toward a unified, modality-agnostic view of in-context learning.