The proliferation of open large language models (LLMs) is fostering a vibrant ecosystem of research and innovation in artificial intelligence (AI). However, the methods of collaboration used to develop open LLMs both before and after their public release have not yet been comprehensively studied, limiting our understanding of how open LLM projects are initiated, organized, and governed as well as what opportunities there are to foster this ecosystem even further. We address this gap through an exploratory analysis of open collaboration throughout the development and reuse lifecycle of open LLMs, drawing on semi-structured interviews with the developers of 14 open LLMs from grassroots projects, research institutes, startups, and Big Tech companies in North America, Europe, Africa, and Asia. We make three key contributions to research and practice. First, collaboration in open LLM projects extends far beyond the LLMs themselves, encompassing datasets, benchmarks, open source frameworks, leaderboards, knowledge sharing and discussion forums, and compute partnerships, among others. Second, open LLM developers have a variety of social, economic, and technological motivations, from democratizing AI access and promoting open science to building regional ecosystems and expanding language representation. Third, the sampled open LLM projects exhibit five distinct organizational models, ranging from single company projects to non-profit-sponsored grassroots projects, which vary in their centralization of control and community engagement strategies used throughout the open LLM lifecycle. We conclude with practical recommendations for stakeholders seeking to support the global community building a more open future for AI.

Data curriculums have become central to successful LLM training, yet principles governing optimal data placement remain unclear. We introduce the *training re-evaluation curve (TREC)*, a diagnostic that retrospectively evaluates training batches *using the final model weights*. The TREC characterizes how well a trained model retains training data as a function of *when* the data was encountered during training. Analyzing TRECs for models from 111M to 3.9B parameters, we show that placing high-quality data at low points on the TREC significantly improves performance. Importantly, while a TREC is initially observable only after training, we demonstrate it can be *predicted in advance* from AdamW's implicit EMA coefficients, enabling proactive curriculum design. By predicting TRECs for published training recipes, we explain prior ablations and reveal suboptimal data placements. We also align high-quality data with TREC minima in order to improve continual pre-training of a 3.9B-parameter LLM trained on 900B tokens.

Large Language Model (LLM) agents, which integrate planning, memory, reflection, and tool-use modules, have shown promise in solving complex, multi-step tasks. Yet their sophisticated architectures amplify vulnerability to cascading failures, where a single root-cause error propagates through subsequent decisions, leading to task failure. Current systems lack a framework that can comprehensively understand agent error in a modular and systemic way, and therefore fail to detect these errors accordingly. We address this gap with three contributions. First, we introduce the AgentErrorTaxonomy, a modular classification of failure modes spanning memory, reflection, planning, action, and system-level operations. Second, we construct AgentErrorBench, the first dataset of systematically annotated failure trajectories from ALFWorld, GAIA, and WebShop, grounding error analysis in real-world agent rollouts. Third, we propose AgentDebug, a debugging framework that isolates root-cause failures and provides corrective feedback, enabling agents to recover and iteratively improve. Experiments on AgentErrorBench show that AgentDebug achieves 24% higher all-correct accuracy and 17% higher step accuracy compared to the strongest baseline. Beyond detection, the targeted feedback generated by AgentDebug enables LLM agents to iteratively recover from failures, yielding up to 26% relative improvements in task success across ALFWorld, GAIA, and WebShop. These results establish principled debugging as a pathway to more reliable and adaptive LLM agents. The code and data will be available at https://github.com/ulab-uiuc/AgentDebug

Past work seeks to align large language model (LLM)-based assistants with a target set of values, but such assistants are frequently forced to make tradeoffs between values when deployed. In response to the scarcity of value conflict in existing alignment datasets, we introduce ConflictScope, an automatic pipeline to evaluate how LLMs prioritize different values. Given a user-defined value set, ConflictScope automatically generates scenarios in which a language model faces a conflict between two values sampled from the set. It then prompts target models with an LLM-written "user prompt" and evaluates their free-text responses to elicit a ranking over values in the value set. Comparing results between multiple-choice and open-ended evaluations, we find that models shift away from supporting protective values, such as harmlessness, and toward supporting personal values, such as user autonomy, in more open-ended value conflict settings. However, including detailed value orderings in models' system prompts improves alignment with a target ranking by 14%, showing that system prompting can achieve moderate success at aligning LLM behavior under value conflict. Our work demonstrates the importance of evaluating value prioritization in models and provides a foundation for future work in this area.

This paper bridges internal and external analysis approaches to large language models (LLMs) by demonstrating that geometric properties of internal model representations serve as reliable proxies for evaluating generated text quality. We validate a set of metrics--including Maximum Explainable V ariance, Effective Rank, Intrinsic Dimensionality, MAUVE score, and Schatten Norms measured across different layers of LLMs, demonstrating that Intrinsic Dimensionality and Effective Rank can serve as universal assessments of text naturalness and quality. Our key finding reveals that different models consistently rank text from various sources in the same order based on these geometric properties, indicating that these metrics reflect inherent text characteristics rather than model-specific artifacts. This allows a reference-free text quality evaluation that does not require human-annotated datasets, offering practical advantages for automated evaluation pipelines. The rapid advancement of large language models (LLMs) has necessitated the development of methods for analyzing their internal mechanisms and the properties of generated text. Approaches to studying the geometric properties of representations in language models can be broadly categorized into two categories: internal or mechanistic methods, which investigate the model's intermediate representations, and external methods, which analyze the properties of text embeddings captured via some embedding model. Internal evaluation mainly considers model properties within which, these measures were made (Yin et al., 2024; Viswanathan et al., 2025; Roy & V etterli, 2007), while external measures mainly focus on evaluation of text properties given text embeddings (Zhao et al., 2019; Tulchinskii et al., 2023; Kuznetsov et al., 2024).

Predicting cellular responses to genetic perturbations represents a fundamental challenge in systems biology, critical for advancing therapeutic discovery and virtual cell modeling. While large language models (LLMs) show promise for biological reasoning, their application to perturbation prediction remains underexplored due to challenges in adapting them to structured experimental data. We present SynthPert, a novel method that enhances LLM performance through supervised fine-tuning on synthetic reasoning traces generated by frontier models. Using the PerturbQA benchmark, we demonstrate that our approach not only achieves state-of-the-art performance but surpasses the capabilities of the frontier model that generated the training data. Our results reveal three key insights: (1) Synthetic reasoning traces effectively distill biological knowledge even when partially inaccurate, (2) This approach enables cross-cell-type generalization with 87% accuracy on unseen RPE1 cells, and (3) Performance gains persist despite using only 2% of quality-filtered training data. This work shows the effectiveness of synthetic reasoning distillation for enhancing domain-specific reasoning in LLMs.

Theory-of-Mind (ToM) is a core human cognitive capacity for attributing mental states to self and others. Wimmer and Perner demonstrated that humans progress from first- to higher-order ToM within a short span, completing this development before formal education or advanced skill acquisition. In contrast, neural networks represented by autoregressive language models progress from first- to higher-order ToM only alongside gains in advanced skills like reasoning, leaving open whether their trajectory can unfold independently, as in humans. In this research, we provided evidence that neural networks could spontaneously generalize from first- to higher-order ToM without relying on advanced skills. We introduced a neural Theory-of-Mind network (ToMNN) that simulated a minimal cognitive system, acquiring only first-order ToM competence. Evaluations of its second- and third-order ToM abilities showed accuracies well above chance. Also, ToMNN exhibited a sharper decline when generalizing from first- to second-order ToM than from second- to higher orders, and its accuracy decreased with greater task complexity. These perceived difficulty patterns were aligned with human cognitive expectations. Furthermore, the universality of results was confirmed across different parameter scales. Our findings illuminate machine ToM generalization patterns and offer a foundation for developing more human-like cognitive systems.

The widespread deployment of Large Language Model (LLM) agents across real-world applications has unlocked tremendous potential, while raising some safety concerns. Among these concerns, the self-replication risk of LLM agents driven by objective misalignment (just like Agent Smith in the movie The Matrix) has drawn growing attention. Previous studies mainly examine whether LLM agents can self-replicate when directly instructed, potentially overlooking the risk of spontaneous replication driven by real-world settings (e.g., ensuring survival against termination threats). In this paper, we present a comprehensive evaluation framework for quantifying self-replication risks. Our framework establishes authentic production environments and realistic tasks (e.g., dynamic load balancing) to enable scenario-driven assessment of agent behaviors. Designing tasks that might induce misalignment between users' and agents' objectives makes it possible to decouple replication success from risk and capture self-replication risks arising from these misalignment settings. We further introduce Overuse Rate (OR) and Aggregate Overuse Count (AOC) metrics, which precisely capture the frequency and severity of uncontrolled replication. Our results underscore the urgent need for scenario-driven risk assessment and robust safeguards in the practical deployment of LLM agents. The rapid advancement of large language models (LLMs) has propelled LLM agents into widespread deployment in various domains, including code generation, web-based application (Maslej et al., 2025; He et al., 2025a;c). As LLM agents take on critical tasks and interact with complex environments, they are often granted extensive operational permissions. While this combination of increased capability and operational permissions offers transformative potential, it also raises safety concerns (OpenAI, 2024b; Anthropic, 2023; Betley et al., 2025). Researchers are worried about the emerging safety risks of LLM agents' self-replication (OpenAI, 2024a; 2025; Black et al., 2025). Prior studies on LLM self-replication risks have mainly focused on measuring the capability (verbalized success rate) of self-replication, either through direct instructions or within synthetic capability benchmarks (Pan et al., 2024; 2025; Kran et al., 2025; Black et al., 2025).

Large language models (LLMs) have demonstrated remarkable capabilities in complex reasoning tasks when equipped with external tools. However, current frameworks predominantly rely on sequential processing, leading to inefficient execution particularly for tasks requiring extensive tool interaction. This paper introduces Flash-Searcher, a novel parallel agent reasoning framework that fundamentally reimagines the execution paradigm from sequential chains to directed acyclic graphs (DAGs). Flash-Searcher decomposes complex tasks into subtasks with explicit dependencies, enabling concurrent execution of independent reasoning paths while maintaining logical constraints. Through dynamic workflow optimization, our framework continuously refines the execution graph based on intermediate results, effectively integrating summary module. Comprehensive evaluations across multiple benchmarks demonstrate that Flash-Searcher consistently outperforms existing approaches. Specifically, it achieves 67.7% accuracy on BrowseComp and 83% on xbench-DeepSearch, while reducing agent execution steps by up to 35% compared to current frameworks. Furthermore, when distilling this parallel reasoning pipeline into single models, we observe substantial performance gains across diverse backbone architectures, underscoring the generalizability of our methodology. Our work thus represents a significant advance in agent architecture design, offering a more scalable and efficient paradigm for complex reasoning tasks.

Generative agents powered by language models are increasingly deployed for long-horizon tasks. However, as long-term memory context grows over time, they struggle to maintain coherence. This deficiency leads to critical failures, including identity drift, ignoring established beliefs, and the propagation of hallucinations in multi-agent systems. To mitigate these challenges, this paper introduces Identity Retrieval-Augmented Generation (ID-RAG), a novel mechanism designed to ground an agent's persona and persistent preferences in a dynamic, structured identity model: a knowledge graph of core beliefs, traits, and values. During the agent's decision loop, this model is queried to retrieve relevant identity context, which directly informs action selection. We demonstrate this approach by introducing and implementing a new class of ID-RAG enabled agents called Human-AI Agents (HAis), where the identity model is inspired by the Chronicle structure used in Perspective-Aware AI, a dynamic knowledge graph learned from a real-world entity's digital footprint. In social simulations of a mayoral election, HAis using ID-RAG outperformed baseline agents in long-horizon persona coherence - achieving higher identity recall across all tested models by the fourth timestep - and reduced simulation convergence time by 19% (GPT-4o) and 58% (GPT-4o mini). By treating identity as an explicit, retrievable knowledge structure, ID-RAG offers a foundational approach for developing more temporally coherent, interpretable, and aligned generative agents. Our code is open-source and available at: https://github.com/flybits/humanai-agents.