Large language models (LLMs) make remarkable progress in reasoning tasks. Among different reasoning modes, inductive reasoning, due to its better alignment with human learning, attracts increasing interest. However, research on inductive reasoning faces certain challenges. First, existing inductive data mostly focuses on superficial regularities while lacking more complex internal patterns. Second, current works merely prompt LLMs or finetune on simple prompt-response pairs, but do not provide precise thinking processes nor implement difficulty control. Unlike previous work, we address these challenges by introducing \textit{CodeSeq}, a synthetic post-training dataset built from number sequences. We package number sequences into algorithmic problems to discover their general terms, defining a general term generation (GTG) task correspondingly. Our pipeline generates supervised finetuning data by reflecting on failed test cases and incorporating iterative corrections, thereby teaching LLMs to learn autonomous case generation and self-checking. Additionally, it leverages reinforcement learning with a novel Case-Synergy Solvability Scaling Reward based on both solvability, estimated from the problem pass rate, and the success rate of self-directed case generation, enabling models to learn more effectively from both successes and failures. Experimental results show that the models trained with \textit{CodeSeq} improve on various reasoning tasks and can preserve the models' OOD performance.

Current Retrieval-Augmented Generation (RAG) systems primarily operate on unimodal textual data, limiting their effectiveness on unstructured multimodal documents. Such documents often combine text, images, tables, equations, and graphs, each contributing unique information. In this work, we present a Modality-Aware Hybrid retrieval Architecture (MAHA), designed specifically for multimodal question answering with reasoning through a modality-aware knowledge graph. MAHA integrates dense vector retrieval with structured graph traversal, where the knowledge graph encodes cross-modal semantics and relationships. This design enables both semantically rich and context-aware retrieval across diverse modalities. Evaluations on multiple benchmark datasets demonstrate that MAHA substantially outperforms baseline methods, achieving a ROUGE-L score of 0.486, providing complete modality coverage. These results highlight MAHA's ability to combine embeddings with explicit document structure, enabling effective multimodal retrieval. Our work establishes a scalable and interpretable retrieval framework that advances RAG systems by enabling modality-aware reasoning over unstructured multimodal data.

Large language models promise a broad set of functions, but when not given a specific objective, they default to milquetoast results such as drafting emails littered with cliches. We demonstrate that inferring the user's in-the-moment objective, then rapidly optimizing for that singular objective, enables LLMs to produce tools, interfaces, and responses that are more responsive and desired. We contribute an architecture for automatically inducing just-in-time objectives by passively observing user behavior, then steering downstream AI systems through generation and evaluation against this objective. Inducing just-in-time objectives (e.g., "Clarify the abstract's research contribution") enables automatic generation of tools, e.g., those that critique a draft based on relevant HCI methodologies, anticipate related researchers' reactions, or surface ambiguous terminology. In a series of experiments (N=14, N=205) on participants' own tasks, JIT objectives enable LLM outputs that achieve 66-86% win rates over typical LLMs, and in-person use sessions (N=17) confirm that JIT objectives produce specialized tools unique to each participant.

Vision-language models have achieved remarkable success in cross-modal understanding. Yet, these models remain limited to object-level or region-level grounding, lacking the capability for pixel-precise keypoint comprehension through natural language. We introduce a novel framework for pixel level grounding. The framework consists of two complementary components: a Point Descriptor that generates rich, contextual descriptions of individual keypoints, and a Point Localizer that regresses precise pixel coordinates from these descriptions. Unlike prior work that relies on templated prompts or keypoint names, our approach produces free-form, coarse-to-fine descriptions that situate keypoints within their visual context. Since there is no available dataset to train such a system, we introduce LlamaPointInPart, a carefully curated dataset of 20K+ image-keypoint-description triplets synthesized from multiple vision-language models, capturing multi-scale information from scene-level context to visual features around the keypoint. For cross-category generalization, we optimize the Point Descriptor on AP-10K via GRPO, using the frozen Point Localizer as a reward model to produce descriptions that maximize localization accuracy. To evaluate our results we establish a new evaluation protocol. Instead of comparing the text description produced by our method to the ground truth, we use the localizer to determine how close is the predicted point generated to the ground truth point. Experiments demonstrate superior performance compared to baseline models on LlamaPointInPart.The bidirectional nature of our framework should enable future applications in both keypoint-guided image understanding and language-guided precise localization. Our code and dataset are publicly available at https://github.com/matanr/Talking_Points.

Obtaining high-quality labels for large datasets is expensive, requiring massive annotations from human experts. While AI models offer a cost-effective alternative by predicting labels, their label quality is compromised by the unavoidable labeling errors. Existing methods mitigate this issue through selective labeling, where AI labels a subset and human labels the remainder. However, these methods lack theoretical guarantees on the quality of AI-assigned labels, often resulting in unacceptably high labeling error within the AI-labeled subset. To address this, we introduce \textbf{Conformal Labeling}, a novel method to identify instances where AI predictions can be provably trusted. This is achieved by controlling the false discovery rate (FDR), the proportion of incorrect labels within the selected subset. In particular, we construct a conformal $p$-value for each test instance by comparing AI models' predicted confidence to those of calibration instances mislabeled by AI models. Then, we select test instances whose $p$-values are below a data-dependent threshold, certifying AI models' predictions as trustworthy. We provide theoretical guarantees that Conformal Labeling controls the FDR below the nominal level, ensuring that a predefined fraction of AI-assigned labels is correct on average. Extensive experiments demonstrate that our method achieves tight FDR control with high power across various tasks, including image and text labeling, and LLM QA.

Reduced-precision data formats are crucial for cost-effective serving of large language models (LLMs). While numerous reduced-precision formats have been introduced thus far, they often require intrusive modifications to the software frameworks or are rather unconventional for widespread adoption across hardware vendors. In this paper, we instead focus on recent industry-driven variants of block floating-point (BFP) formats and conduct a comprehensive analysis to push their limits for efficient LLM serving. Our analysis shows that existing ultra low-bit BFP variants struggle to provide reasonable language model performance due to outlier values in blocks. To address the outliers with BFPs, we propose MX+, a cost-effective and non-intrusive extension designed for seamless integration into the microscaling (MX) formats. MX+ builds on the key insight that the outlier does not need to use its exponent field in the element data type, which allows us to repurpose the exponent field as an extended mantissa to increase the precision of the outlier element. Our evaluation shows that MX+ achieves significantly higher model performance compared to the 4-bit MX format (MXFP4) with negligible storage overhead and slowdown, thus offering a compelling alternative to MXFP4 or MXFP6 for efficient LLM inference.

Recent LLM agents have made great use of chain of thought reasoning and function calling. As their capabilities grow, an important question arises: can this software represent not only a smart problem-solving tool, but an entity in its own right, that can plan, design immediate tasks, and reason toward broader, more ambiguous goals? To study this question, we adopt an open-ended experimental setting where we augment a pretrained LLM agent with the ability to generate its own tasks, accumulate knowledge, and interact extensively with its environment. We study the resulting open-ended agent qualitatively. It can reliably follow complex multi-step instructions, store and reuse information across runs, and propose and solve its own tasks, though it remains sensitive to prompt design, prone to repetitive task generation, and unable to form self-representations. These findings illustrate both the promise and current limits of adapting pretrained LLMs toward open-endedness, and point to future directions for training agents to manage memory, explore productively, and pursue abstract long-term goals.

Embeddings from Visual-Language Models are increasingly utilized to represent semantics in robotic maps, offering an open-vocabulary scene understanding that surpasses traditional, limited labels. Embeddings enable on-demand querying by comparing embedded user text prompts to map embeddings via a similarity metric. The key challenge in performing the task indicated in a query is that the robot must determine the parts of the environment relevant to the query. This paper proposes a solution to this challenge. We leverage natural-language synonyms and antonyms associated with the query within the embedding space, applying heuristics to estimate the language space relevant to the query, and use that to train a classifier to partition the environment into matches and non-matches. We evaluate our method through extensive experiments, querying both maps and standard image benchmarks. The results demonstrate increased queryability of maps and images. Our querying technique is agnostic to the representation and encoder used, and requires limited training.

Recently, Agentic Reinforcement Learning (Agentic RL) has made significant progress in incentivizing the multi-turn, long-horizon tool-use capabilities of web agents. While mainstream agentic RL algorithms autonomously explore high-uncertainty tool-call steps under the guidance of entropy, excessive reliance on entropy signals can impose further constraints, leading to the training collapse. In this paper, we delve into the challenges caused by entropy and propose the Agentic Entropy-Balanced Policy Optimization (AEPO), an agentic RL algorithm designed to balance entropy in both the rollout and policy update phases. AEPO comprises two core components: (1) a dynamic entropy-balanced rollout mechanism that adaptively allocate global and branch sampling budget through entropy pre-monitoring, while imposing a branch penalty on consecutive high-entropy tool-call steps to prevent over-branching issues; and (2) Entropy-Balanced Policy Optimization that inserts a stop-gradient operation into the high-entropy clipping term to preserve and properly rescale gradients on high-entropy tokens, while incorporating entropy-aware advantage estimation to prioritize learning on high-uncertainty tokens. Results across 14 challenging datasets show that AEPO consistently outperforms 7 mainstream RL algorithms. With just 1K RL samples, Qwen3-14B with AEPO achieves impressive results: 47.6% on GAIA, 11.2% on Humanity's Last Exam, and 43.0% on WebWalker for Pass@1; 65.0% on GAIA, 26.0% on Humanity's Last Exam, and 70.0% on WebWalker for Pass@5. Further analysis reveals that AEPO improves rollout sampling diversity while maintaining stable policy entropy, facilitating scalable web agent training.

AI systems are continually evolving and advancing, and user expectations are concurrently increasing, with a growing demand for interactions that go beyond simple text-based interaction with Large Language Models (LLMs). Today's applications often require LLMs to interact with external tools, marking a shift toward more complex agentic systems. To support this, standards such as the Model Context Protocol (MCP) have emerged, enabling agents to access tools by including a specification of the capabilities of each tool within the prompt. Although this approach expands what agents can do, it also introduces a growing problem: prompt bloating. As the number of tools increases, the prompts become longer, leading to high prompt token costs, increased latency, and reduced task success resulting from the selection of tools irrelevant to the prompt. To address this issue, we introduce JSPLIT, a taxonomy-driven framework designed to help agents manage prompt size more effectively when using large sets of MCP tools. JSPLIT organizes the tools into a hierarchical taxonomy and uses the user's prompt to identify and include only the most relevant tools, based on both the query and the taxonomy structure. In this paper, we describe the design of the taxonomy, the tool selection algorithm, and the dataset used to evaluate JSPLIT. Our results show that JSPLIT significantly reduces prompt size without significantly compromising the agent's ability to respond effectively. As the number of available tools for the agent grows substantially, JSPLIT even improves the tool selection accuracy of the agent, effectively reducing costs while simultaneously improving task success in high-complexity agent environments.