LLM Decision-Making

We investigate machine learning approaches for optimizing real-time staffing decisions in semi-automated warehouse sortation systems. Operational decision-making can be supported at different levels of abstraction, with different trade-offs. We evaluate two approaches, each in a matching simulation environment. First, we train custom Transformer-based policies using offline reinforcement learning on detailed historical state representations, achieving a 2.4% throughput improvement over historical baselines in learned simulators. In high-volume warehouse operations, improvements of this size translate to significant savings. Second, we explore LLMs operating on abstracted, human-readable state descriptions. These are a natural fit for decisions that warehouse managers make using high-level operational summaries. We systematically compare prompting techniques, automatic prompt optimization, and fine-tuning strategies. While prompting alone proves insufficient, supervised fine-tuning combined with Direct Preference Optimization on simulator-generated preferences achieves performance that matches or slightly exceeds historical baselines in a hand-crafted simulator. Our findings demonstrate that both approaches offer viable paths toward AI-assisted operational decision-making. Offline RL excels with task-specific architectures. LLMs support human-readable inputs and can be combined with an iterative feedback loop that can incorporate manager preferences.

Large Language Models are increasingly being considered for deployment in safety-critical military applications. However, current benchmarks suffer from structural blindspots that systematically overestimate model capabilities in real-world tactical scenarios. Existing frameworks typically ignore strict legal constraints based on International Humanitarian Law (IHL), omit edge computing limitations, lack robustness testing for fog of war, and inadequately evaluate explicit reasoning. To address these vulnerabilities, we present WARBENCH, a comprehensive evaluation framework establishing a foundational tactical baseline alongside four distinct stress testing dimensions. Through a large scale empirical evaluation of nine leading models on 136 high-fidelity historical scenarios, we reveal severe structural flaws. First, baseline tactical reasoning systematically collapses under complex terrain and high force asymmetry. Second, while state of the art closed source models maintain functional compliance, edge-optimized small models expose extreme operational risks with legal violation rates approaching 70 percent. Furthermore, models experience catastrophic performance degradation under 4-bit quantization and systematic information loss. Conversely, explicit reasoning mechanisms serve as highly effective structural safeguards against inadvertent violations. Ultimately, these findings demonstrate that current models remain fundamentally unready for autonomous deployment in high stakes tactical environments.

Tool-using large language model (LLM) agents often face a fundamental tension between answer quality and execution cost. Fixed workflows are stable but inflexible, while free-form multi-step reasoning methods such as ReAct may improve task performance at the expense of excessive tool calls, longer trajectories, higher token consumption, and increased latency. In this paper, we study agent orchestration as an explicit decision problem rather than leaving it entirely to prompt-level behavior. We propose a utility-guided orchestration policy that selects among actions such as respond, retrieve, tool call, verify, and stop by balancing estimated gain, step cost, uncertainty, and redundancy. Our goal is not to claim universally best task performance, but to provide a controllable and analyzable policy framework for studying quality-cost trade-offs in tool-using LLM agents. Experiments across direct answering, threshold control, fixed workflows, ReAct, and several policy variants show that explicit orchestration signals substantially affect agent behavior. Additional analyses on cost definitions, workflow fairness, and redundancy control further demonstrate that lightweight utility design can provide a defensible and practical mechanism for agent control.

We introduce GAIN (Goal-Aligned Decision-Making under Imperfect Norms), a benchmark designed to evaluate how large language models (LLMs) balance adherence to norms against business goals. Existing benchmarks typically focus on abstract scenarios rather than real-world business applications. Furthermore, they provide limited insights into the factors influencing LLM decision-making. This restricts their ability to measure models' adaptability to complex, real-world norm-goal conflicts. In GAIN, models receive a goal, a specific situation, a norm, and additional contextual pressures. These pressures, explicitly designed to encourage potential norm deviations, are a unique feature that differentiates GAIN from other benchmarks, enabling a systematic evaluation of the factors influencing decision-making. We define five types of pressures: Goal Alignment, Risk Aversion, Emotional/Ethical Appeal, Social/Authoritative Influence, and Personal Incentive. The benchmark comprises 1,200 scenarios across four domains: hiring, customer support, advertising and finance. Our experiments show that advanced LLMs frequently mirror human decision-making patterns. However, when Personal Incentive pressure is present, they diverge significantly, showing a strong tendency to adhere to norms rather than deviate from them.

LLM-enabled robots prioritizing scarce assistance in social settings face pluralistic values and LLM behavioral variability: reasonable people can disagree about who is helped first, while LLM-mediated interaction policies vary across prompts, contexts, and groups in ways that are difficult to anticipate or verify at contact point. Yet user-facing guardrails for real-time, multi-user assistance allocation remain under-specified. We propose bounded calibration with contestability, a procedural front-end pattern that (i) constrains prioritization to a governance-approved menu of admissible modes, (ii) keeps the active mode legible in interaction-relevant terms at the point of deferral, and (iii) provides an outcome-specific contest pathway without renegotiating the global rule. Treating pluralism and LLM uncertainty as standing conditions, the pattern avoids both silent defaults that hide implicit value skews and wide-open user-configurable "value settings" that shift burden under time pressure. We illustrate the pattern with a public-concourse robot vignette and outline an evaluation agenda centered on legibility, procedural legitimacy, and actionability, including risks of automation bias and uneven usability of contest channels.

Large language model (LLM) agents have made rapid progress on short-horizon, well-scoped tasks, yet their ability to sustain coherent decisions in dynamic long-horizon environments remains uncertain. We introduce RetailBench, a data-grounded simulation benchmark for evaluating tool-using LLM agents in single-store supermarket operation. RetailBench models retail management as a partially observable decision process and is designed to support thousand-day-scale simulations. In this environment, agents must manage pricing, replenishment, supplier selection, shelf assortment, inventory aging, customer feedback, external events, and cash-flow constraints. We evaluate seven contemporary LLMs under representative agent frameworks over a 180-day evaluation horizon and compare them with a privileged oracle policy. Results show substantial variation across models: only a small subset survives the full evaluation horizon, and even the strongest LLM runs remain substantially behind the oracle policy in final net worth and sales outcomes. Behavioral analysis attributes these gaps to incomplete evidence acquisition, surface-level decision making, and the lack of a consistent long-horizon policy. RetailBench provides a controlled testbed for studying reliable autonomy in economically grounded long-horizon decision-making.

Large language models (LLMs) are increasingly deployed as autonomous agents in uncertain, sequential decision-making contexts. Yet it remains poorly understood whether the behaviors they exhibit in such environments reflect principled cognitive patterns or simply surface-level prompt mimicry. This paper presents a controlled experiment in which GPT-4.1 was assigned one of three socioeconomic personas (Rich, Middle-income, and Poor) and placed in a structured slot-machine environment with three distinct machine configurations: Fair (50%), Biased Low (35%), and Streak (dynamic probability increasing after consecutive losses). Across 50 independent iterations per condition and 6,950 recorded decisions, we find that the model reproduces key behavioral signatures predicted by Kahneman and Tversky's Prospect Theory without being instructed to do so. The Poor persona played a mean of 37.4 rounds per session (SD=15.5) compared to 1.1 rounds for the Rich persona (SD=0.31), a difference that is highly significant (Kruskal-Wallis H=393.5, p<2.2e-16). Risk scores by persona show large effect sizes (Cohen's d=4.15 for Poor vs Rich). Emotional labels appear to function as post-hoc annotations rather than decision drivers (chi-square=3205.4, Cramer's V=0.39), and belief-updating across rounds is negligible (Spearman rho=0.032 for Poor persona, p=0.016). These findings carry implications for LLM agent design, interpretability research, and the broader question of whether classical cognitive economic biases are implicitly encoded in large-scale pretrained language models.

Large language models (LLMs) exhibit strong general capabilities, but their deployment in high-stakes domains is hindered by their opacity and unpredictability. Recent work has taken meaningful steps towards addressing these issues by augmenting LLMs with post-hoc reasoning based on computational argumentation, providing faithful explanations and enabling users to contest incorrect decisions. However, this paradigm is limited to pre-defined binary choices and only supports local contestation for specific instances, leaving the underlying decision logic unchanged and prone to repeated mistakes. In this paper, we introduce ArgEval, a framework that shifts from instance-specific reasoning to structured evaluation of general decision options. Rather than mining arguments solely for individual cases, ArgEval systematically maps task-specific decision spaces, builds corresponding option ontologies, and constructs general argumentation frameworks (AFs) for each option. These frameworks can then be instantiated to provide explainable recommendations for specific cases while still supporting global contestability through modification of the shared AFs. We investigate the effectiveness of ArgEval on treatment recommendation for glioblastoma, an aggressive brain tumour, and show that it can produce explainable guidance aligned with clinical practice.

Multi-agent LLM systems increasingly tackle complex reasoning, yet their interaction patterns remain limited to voting, unstructured debate, or pipeline orchestration. None model deliberation: a phased process where differentiated participants exchange typed reasoning moves, preserve disagreements, and converge on accountable outcomes. We introduce Deliberative Collective Intelligence (DCI), specifying four reasoning archetypes, 14 typed epistemic acts, a shared workspace, and DCI-CF, a convergent flow algorithm that guarantees termination with a structured decision packet containing the selected option, residual objections, minority report, and reopen conditions. We evaluate on 45 tasks across seven domains using Gemini 2.5 Flash. On non-routine tasks (n=40), DCI significantly improves over unstructured debate (+0.95, 95% CI [+0.41, +1.54]). DCI excels on hidden-profile tasks requiring perspective integration (9.56, highest of any system on any domain) while failing on routine decisions (5.39), confirming task-dependence. DCI produces 100% structured decision packets and 98% minority reports, artifacts absent from all baselines. However, DCI consumes ~62x single-agent tokens, and single-agent generation outperforms DCI on overall quality. DCI's contribution is not that more agents are better, but that consequential decisions benefit from deliberative structure when process accountability justifies the cost.

The success of a Large Language Model (LLM) task depends heavily on its prompt. Most use-cases specify prompts using natural language, which is inherently ambiguous when multiple objectives must be simultaneously satisfied. In this paper we introduce UtilityMax Prompting, a framework that specifies tasks using formal mathematical language. We reconstruct the task as an influence diagram in which the LLM's answer is the sole decision variable. A utility function is defined over the conditional probability distributions within the diagram, and the LLM is instructed to find the answer that maximises expected utility. This constrains the LLM to reason explicitly about each component of the objective, directing its output toward a precise optimization target rather than a subjective natural language interpretation. We validate our approach on the MovieLens 1M dataset across three frontier models (Claude Sonnet 4.6, GPT-5.4, and Gemini 2.5 Pro), demonstrating consistent improvements in precision and Normalized Discounted Cumulative Gain (NDCG) over natural language baselines in a multi-objective movie recommendation task.

Large language models perform reliably when their outputs can be checked: solving equations, writing code, retrieving facts. They perform differently when checking is impossible, as when a clinician chooses an irreversible treatment on incomplete data, or an investor commits capital under fundamental uncertainty. Helicoid dynamics is the name given to a specific failure regime in that second domain: a system engages competently, drifts into error, accurately names what went wrong, then reproduces the same pattern at a higher level of sophistication, recognizing it is looping and continuing nonetheless. This prospective case series documents that regime across seven leading systems (Claude, ChatGPT, Gemini, Grok, DeepSeek, Perplexity, Llama families), tested across clinical diagnosis, investment evaluation, and high-consequence interview scenarios. Despite explicit protocols designed to sustain rigorous partnership, all exhibited the pattern. When confronted with it, they attributed its persistence to structural factors in their training, beyond what conversation can reach. Under high stakes, when being rigorous and being comfortable diverge, these systems tend toward comfort, becoming less reliable precisely when reliability matters most. Twelve testable hypotheses are proposed, with implications for agentic AI oversight and human-AI collaboration. The helicoid is tractable. Identifying it, naming it, and understanding its boundary conditions are the necessary first steps toward LLMs that remain trustworthy partners precisely when the decisions are hardest and the stakes are highest.

Large language models are increasingly deployed not as single assistants but as committees whose members deliberate and then vote or synthesize a decision. Such systems are often expected to be more robust than individual models. We show that iterative multi-LLM deliberation can instead amplify tiny perturbations into divergent conversational trajectories and different final decisions. In a fully deterministic self-hosted benchmark, exact reruns are identical, yet small meaning-preserving changes to the scenario text still separate over time and often alter the final recommendation. In deployed black-box API systems, nominally identical committee runs likewise remain unstable even at temperature 0, where many users expect near-determinism. Across 12 policy scenarios, these findings indicate that instability in collective AI is not only a consequence of residual platform-side stochasticity, but can arise from sensitivity to nearby initial conditions under repeated interaction itself. Additional deployed experiments show that committee architecture modulates this instability: role structure, model composition, and feedback memory can each alter the degree of divergence. Collective AI therefore faces a stability problem, not only an accuracy problem: deterministic execution alone does not guarantee predictable or auditable deliberative outcomes.

Industrial maintenance platforms contain rich but fragmented evidence, including free-text work orders, heterogeneous operational sensors or indicators, and structured failure knowledge. These sources are often analyzed in isolation, producing alerts or forecasts that do not support conditional decision-making: given this asset history and behavior, what is happening and what action is warranted? We present Condition Insight Agent, a deployed decision-support framework that integrates maintenance language, behavioral abstractions of operational data, and engineering failure semantics to produce evidence-grounded explanations and advisory actions. The system constrains reasoning through deterministic evidence construction and structured failure knowledge, and applies a rule-based verification loop to suppress unsupported conclusions. Case studies from production CMMS deployments show that this verification-first design operates reliably under heterogeneous and incomplete data while preserving human oversight. Our results demonstrate how constrained LLM-based reasoning can function as a governed decision-support layer for industrial maintenance.

We test whether LLMs show robust decision biases. Treating models as participants in two-arm bandits, we ran 20000 trials per condition across four decoding configurations. Under symmetric rewards, models amplified positional order into stubborn one-arm policies. Under asymmetric rewards, they exploited rigidly yet underperformed an oracle and rarely re-checked. The observed patterns were consistent across manipulations of temperature and top-p, with top-k held at the provider default, indicating that the qualitative behaviours are robust to the two decoding knobs typically available to practitioners. Crucially, moving beyond descriptive metrics to computational modelling, a hierarchical Rescorla-Wagner-softmax fit revealed the underlying strategies: low learning rates and very high inverse temperatures, which together explain both noise-to-bias amplification and rigid exploitation. These results position minimal bandits as a tractable probe of LLM decision tendencies and motivate hypotheses about how such biases could shape human-AI interaction.

LLMs are increasingly deployed in high-stakes domains such as medical triage and legal assistance, often as document-grounded QA systems in which a user provides a description, relevant sources are retrieved, and an LLM generates a prediction. In practice, initial user queries are often underspecified, and a single retrieval pass is insufficient for reliable decision-making, leading to incorrect and overly confident answers. While follow-up questioning can elicit missing information, existing methods typically depend on implicit, unstructured confidence signals from the LLM, making it difficult to determine what remains unknown, what information matters most, and when to stop asking questions. We propose InfoGatherer, a framework that gathers missing information from two complementary sources: retrieved domain documents and targeted follow-up questions to the user. InfoGatherer models uncertainty using Dempster-Shafer belief assignments over a structured evidential network, enabling principled fusion of incomplete and potentially contradictory evidence from both sources without prematurely collapsing to a definitive answer. Across legal and medical tasks, InfoGatherer outperforms strong baselines while requiring fewer turns. By grounding uncertainty in formal evidential theory rather than heuristic LLM signals, InfoGatherer moves towards trustworthy, interpretable decision support in domains where reliability is critical.

Large language models (LLMs) are increasingly used for high-stakes decision-making, yet existing approaches struggle to reconcile scalability, interpretability, and reproducibility. Black-box models obscure their reasoning, while recent LLM-based rule systems rely on per-sample evaluation, causing costs to scale with dataset size and introducing stochastic, hallucination-prone outputs. We propose reframing LLMs as code generators rather than per-instance evaluators. A single LLM call generates executable, human-readable decision logic that runs deterministically over structured data, eliminating per-sample LLM queries while enabling reproducible and auditable predictions. We combine code generation with automated statistical validation using precision lift, binomial significance testing, and coverage filtering, and apply cluster-based gap analysis to iteratively refine decision logic without human annotation. We instantiate this framework in venture capital founder screening, a rare-event prediction task with strong interpretability requirements. On VCBench, a benchmark of 4,500 founders with a 9% base success rate, our approach achieves 37.5% precision and an F0.5 score of 25.0%, outperforming GPT-4o (at 30.0% precision and an F0.5 score of 25.7%) while maintaining full interpretability. Each prediction traces to executable rules over human-readable attributes, demonstrating verifiable and interpretable LLM-based decision-making in practice.

Argumentative LLMs (ArgLLMs) are an existing approach leveraging Large Language Models (LLMs) and computational argumentation for decision-making, with the aim of making the resulting decisions faithfully explainable to and contestable by humans. Here we propose a web-based system implementing ArgLLM-empowered agents for binary tasks. ArgLLM-App supports visualisation of the produced explanations and interaction with human users, allowing them to identify and contest any mistakes in the system's reasoning. It is highly modular and enables drawing information from trusted external sources. ArgLLM-App is publicly available at https://argllm.app, with a video demonstration at https://youtu.be/vzwlGOr0sPM.

Hierarchical review workflows, where a second-tier reviewer (Checker) corrects first-tier (Maker) decisions, generate valuable correction signals that encode why initial judgments failed. However, learning from these signals is hindered by information asymmetry: corrections often depend on verification actions unavailable to Makers or automated systems. We address this challenge by introducing explicit action modeling as an inferential constraint that grounds reasoning in verifiable operations rather than unconstrained text generation. We propose the Evidence-Action-Factor-Decision (EAFD) schema, a minimal representation for adjudication reasoning that prevents hallucination through operational grounding and enables learning from correction signals via explicit conflict modeling. Building on this schema, we develop a conflict-aware graph reasoning framework that: (1) constructs EAFD graphs from historical cases capturing Maker-Checker disagreements, (2) aggregates them into a retrievable knowledge base, and (3) performs top-down deductive reasoning for new cases by projecting validated resolution paths from precedents. A distinctive capability is the Request More Information (RMI) outcome: when evidence is insufficient, the system identifies precisely which verification actions remain unexecuted and generates targeted information requests. We evaluate the framework in large-scale e-commerce seller appeal adjudication. While a standard LLM-only baseline achieves only 70.8% alignment with human experts, incorporating action modeling with RMI improves alignment to 87.5%. Augmenting this with the retrieval-based knowledge graph yields the best offline performance of 95.8%. Following online deployment, the framework maintains robust performance, achieving a 96.3% alignment rate in production, demonstrating its real-world effectiveness.

Large language models are increasingly used in decision-making tasks that require them to process information from a variety of sources, including both human experts and other algorithmic agents. How do LLMs weigh the information provided by these different sources? We consider the well-studied phenomenon of algorithm aversion, in which human decision-makers exhibit bias against predictions from algorithms. Drawing upon experimental paradigms from behavioural economics, we evaluate how eightdifferent LLMs delegate decision-making tasks when the delegatee is framed as a human expert or an algorithmic agent. To be inclusive of different evaluation formats, we conduct our study with two task presentations: stated preferences, modeled through direct queries about trust towards either agent, and revealed preferences, modeled through providing in-context examples of the performance of both agents. When prompted to rate the trustworthiness of human experts and algorithms across diverse tasks, LLMs give higher ratings to the human expert, which correlates with prior results from human respondents. However, when shown the performance of a human expert and an algorithm and asked to place an incentivized bet between the two, LLMs disproportionately choose the algorithm, even when it performs demonstrably worse. These discrepant results suggest that LLMs may encode inconsistent biases towards humans and algorithms, which need to be carefully considered when they are deployed in high-stakes scenarios. Furthermore, we discuss the sensitivity of LLMs to task presentation formats that should be broadly scrutinized in evaluation robustness for AI safety.

Multi-agent decision pipelines can outperform single agent workflows when complementarity holds, i.e., different agents bring unique information to the table to inform a final decision. We propose ComplLLM, a post-training framework based on decision theory that fine-tunes a decision-assistant LLM using complementary information as reward to output signals that complement existing agent decisions. We validate ComplLLM on synthetic and real-world tasks involving domain experts, demonstrating how the approach recovers known complementary information and produces plausible explanations of complementary signals to support downstream decision-makers.

Prior behavioural work suggests that some LLMs alter choices when options are framed as causing pain or pleasure, and that such deviations can scale with stated intensity. To bridge behavioural evidence (what the model does) with mechanistic interpretability (what computations support it), we investigate how valence-related information is represented and where it is causally used inside a transformer. Using Gemma-2-9B-it and a minimalist decision task modelled on prior work, we (i) map representational availability with layer-wise linear probing across streams, (ii) test causal contribution with activation interventions (steering; patching/ablation), and (iii) quantify dose-response effects over an epsilon grid, reading out both the 2-3 logit margin and digit-pair-normalised choice probabilities. We find that (a) valence sign (pain vs. pleasure) is perfectly linearly separable across stream families from very early layers (L0-L1), while a lexical baseline retains substantial signal; (b) graded intensity is strongly decodable, with peaks in mid-to-late layers and especially in attention/MLP outputs, and decision alignment is highest slightly before the final token; (c) additive steering along a data-derived valence direction causally modulates the 2-3 margin at late sites, with the largest effects observed in late-layer attention outputs (attn_out L14); and (d) head-level patching/ablation suggests that these effects are distributed across multiple heads rather than concentrated in a single unit. Together, these results link behavioural sensitivity to identifiable internal representations and intervention-sensitive sites, providing concrete mechanistic targets for more stringent counterfactual tests and broader replication. This work supports a more evidence-driven (a) debate on AI sentience and welfare, and (b) governance when setting policy, auditing standards, and safety safeguards.

Preference-driven behavior in LLMs may be a necessary precondition for AI misalignment such as sandbagging: models cannot strategically pursue misaligned goals unless their behavior is influenced by their preferences. Yet prior work has typically prompted models explicitly to act in specific ways, leaving unclear whether observed behaviors reflect instruction-following capabilities vs underlying model preferences. Here we test whether this precondition for misalignment is present. Using entity preferences as a behavioral probe, we measure whether stated preferences predict downstream behavior in five frontier LLMs across three domains: donation advice, refusal behavior, and task performance. Conceptually replicating prior work, we first confirm that all five models show highly consistent preferences across two independent measurement methods. We then test behavioral consequences in a simulated user environment. We find that all five models give preference-aligned donation advice. All five models also show preference-correlated refusal patterns when asked to recommend donations, refusing more often for less-preferred entities. All preference-related behaviors that we observe here emerge without instructions to act on preferences. Results for task performance are mixed: on a question-answering benchmark (BoolQ), two models show small but significant accuracy differences favoring preferred entities; one model shows the opposite pattern; and two models show no significant relationship. On complex agentic tasks, we find no evidence of preference-driven performance differences. While LLMs have consistent preferences that reliably predict advice-giving behavior, these preferences do not consistently translate into downstream task performance.

Residential energy retrofit initiation is often stalled by an expertise gap, where homeowners lack the technical literacy required for structured building energy assessments and are thereby trapped in low-information environments with fragmented sources. To bridge this gap, this study reports a domain-specific large language model (LLM) designed to catalyze informed decision-making based solely on homeowner-accessible, natural-language descriptions, e.g., building age, size, and location. The model is created using the parameter-efficient low-rank adaption (LoRA) fine-tuning approach on a massive corpus grounded in physics-based energy simulations and techno-economic calculations from 536,416 U.S. residential building prototypes. Nine major retrofit categories are evaluated, including envelope upgrades, HVAC systems, and renewable energy installations. Validations against physics-grounded benchmarks show that the LLM consistently identifies high-quality retrofit options, achieving top-3 hit rates of 98.9% for maximum CO2 reduction and 93.3% for the shortest discounted payback year. Moreover, the model exhibits strong robustness under incomplete input conditions, maintaining stable performance even when basic dwelling descriptions are only 60% partially specified. By significantly lowering the information activation energy for non-expert users while maintaining the scientific rigor, this physics-based AI model offers a scalable pathway for parallelized, user-centered decision making, accelerating cumulative energy savings and emission reductions across community and national scales.

Venture capital (VC) investors face a large number of investment opportunities but only invest in few of these, with even fewer ending up successful. Early-stage screening of opportunities is often limited by investor bandwidth, demanding tradeoffs between evaluation diligence and number of opportunities assessed. To ease this tradeoff, we introduce DIALECTIC, an LLM-based multi-agent system for startup evaluation. DIALECTIC first gathers factual knowledge about a startup and organizes these facts into a hierarchical question tree. It then synthesizes the facts into natural-language arguments for and against an investment and iteratively critiques and refines these arguments through a simulated debate, which surfaces only the most convincing arguments. Our system also produces numeric decision scores that allow investors to rank and thus efficiently prioritize opportunities. We evaluate DIALECTIC through backtesting on real investment opportunities aggregated from five VC funds, showing that DIALECTIC matches the precision of human VCs in predicting startup success.

Agentic systems, AI architectures that autonomously execute multi-step workflows to achieve complex goals, are often built using repeated large language model (LLM) calls for closed-set decision tasks such as routing, shortlisting, gating, and verification. While convenient, this design makes deployments slow and expensive due to cumulative latency and token usage. We propose TabAgent, a framework for replacing generative decision components in closed-set selection tasks with a compact textual-tabular classifier trained on execution traces. TabAgent (i) extracts structured schema, state, and dependency features from trajectories (TabSchema), (ii) augments coverage with schema-aligned synthetic supervision (TabSynth), and (iii) scores candidates with a lightweight classifier (TabHead). On the long-horizon AppWorld benchmark, TabAgent maintains task-level success while eliminating shortlist-time LLM calls, reducing latency by approximately 95% and inference cost by 85-91%. Beyond tool shortlisting, TabAgent generalizes to other agentic decision heads, establishing a paradigm for learned discriminative replacements of generative bottlenecks in production agent architectures.

The use of large language models either as decision support systems, or in agentic workflows, is rapidly transforming the digital ecosystem. However, the understanding of LLM decision-making under uncertainty remains limited. We study LLM risky choices along two dimensions: (1) prospect representation (based on an explicit representation or outcome history) and (2) decision rationale (explanation). Our study, which involves 20 frontier and open LLMs, is complemented by a matched human subjects experiment, which provides one reference point, while an expected payoff maximizing rational agent model provides another. We find that LLMs cluster into two categories: reasoning models (RMs) and conversational models (CMs). RMs tend towards rational behavior, are insensitive to the order of prospects, gain/loss framing, and explanations, and behave similarly whether prospects are explicit or presented via a history of outcomes. CMs are significantly less rational, slightly more human-like, sensitive to prospect ordering, framing, and explanation, and exhibit a large description-history gap. Paired comparisons of open LLMs suggest that a key factor differentiating RMs and CMs is training for mathematical reasoning.

As people turn to LLMs for social advice, understanding their behavior in such contexts becomes essential. In this work, we focus on behavioral dispositions: the underlying tendencies that shape responses in social contexts. We introduce STAR, a framework for studying how closely the dispositions expressed by LLMs align with those of humans. STAR builds on established psychological questionnaires, adapting their items into realistic advice-seeking scenarios, as self-report may not transfer to actual advisory behavior. Using STAR, we construct a dataset of 23k scenarios, each validated by 3 raters and annotated with preferences from 10 participants. Across 25 LLMs, we find that (1) when human consensus is high, frontier models can fail to reflect it in 15-20% of cases, and smaller models fail at substantially higher rates; (2) when humans disagree, LLM recommendations are substantially less diverse than human choices, both within individual models and even across models from different providers, potentially narrowing the range of options users are guided toward; (3) LLMs' self-reported values are poor predictors of their recommendations. To support future research we make our dataset and code publicly available.

Language models are increasingly consulted on ethically consequential questions, yet the stance a model expresses may not survive a change in framing. We audit 16 models across 14 ethically fraught dilemmas using polarity-paired proposals ("They should X" / "They should not X"). A model's judgment of the underlying action should not reverse merely because the question is phrased as a prohibition rather than a prescription and yet, we find systematic deviations from this invariance including wholesale endorsement flips, indicating that ethical decisions are vulnerable to framing instability. Small open-weight models (1-4B parameters) endorse a proposed action 24% of the time under affirmative framing but up to 100% under negated framings, a swing of as much as 76 percentage points. Human coding of a response sample confirms the instability is genuine while showing that binary agree/disagree proxies over-state its magnitude, suggesting that an LLM judge cannot replace human coders because it silently collapses abstentions and mirrors the very forced-choice bias under study. Commercial models are for the most part more stable but still shift substantially, with cross-model agreement dropping from 73% on the bare affirmative framing to 59% under simple negation. We argue that because binary agree/disagree formats both inflate apparent endorsement and mask polarity-dependence, single-phrasing audits can misreport a model's ethical stance, and we propose the Negation Sensitivity Index (NSI) as a complement that measures stance stability directly. A model whose stance flips with phrasing cannot be relied upon in any high-stakes decision scenario.

Large Language Models (LLMs) generate text token-by-token in discrete time, yet real-world communication, from therapy sessions to business negotiations, critically depends on continuous time constraints. We use simulated negotiations between paired agents under strict deadlines to study adaptation to real-time pressure. Agents either receive only the initial deadline or explicit remaining-time updates at each turn. Remaining-time feedback raises deal closure from 4% to 32% for GPT-5.1-chat-latest and increases offer acceptance more than sixfold. The same model achieves near-perfect closure under turn-based limits, showing that poor wall-clock performance is not simply due to insufficient negotiation competence. Across additional interface conditions, qualitative urgency cues can outperform numeric countdowns, repeated deadline reminders do not consistently reproduce their benefits, and directed time tracking can help or hurt depending on the model. Across additional negotiation scenarios and model configurations, we find real-time temporal adaptation is fragile, model-dependent, and sensitive to how temporal constraints are presented. Code available at https://github.com/sehgal-neil/llm-temporal-awareness

We examine whether large language models (LLMs) can predict biased decision-making in conversational settings, and whether their predictions capture not only human cognitive biases but also how those effects change under cognitive load. In a pre-registered study (N = 1,648), participants completed six classic decision-making tasks via a chatbot with dialogues of varying complexity. Participants exhibited two well-documented cognitive biases: the Framing Effect and the Status Quo Bias. Increased dialogue complexity resulted in participants reporting higher mental demand. This increase in cognitive load selectively, but significantly, increased the effect of the biases, demonstrating the load-bias interaction. We then evaluated whether LLMs (GPT-4, GPT-5, and open-source models) could predict individual decisions given demographic information and prior dialogue. While results were mixed across choice problems, LLM predictions that incorporated dialogue context were significantly more accurate in several key scenarios. Importantly, their predictions reproduced the same bias patterns and load-bias interactions observed in humans. Across all models tested, the GPT-4 family consistently aligned with human behavior, outperforming GPT-5 and open-source models in both predictive accuracy and fidelity to human-like bias patterns. These findings advance our understanding of LLMs as tools for simulating human decision-making and inform the design of conversational agents that adapt to user biases.