Md Khairul Islam, Zeyu Xia, Adam S. Bolton, Judy Fox
Standard sequence models encode position by token rank rather than physical coordinate, typically forcing practitioners to destructively resample continuous physical signals such as spectra observed across heterogeneous instruments. We identify this coordinate mismatch as a structural limitation in cross-instrument representation learning, and propose OmniSpectra, a wavelength-native foundation model for spectral representation learning: a continuous-coordinate masked autoencoder that discards the integer pixel indexing of sequence modeling. Mapping physical coordinates to a logarithmic basis turns the multiplicative scaling induced by cosmological redshift into a uniform additive shift. In continuous rotary position embeddings, this shared shift cancels from relative token distances rather than producing grid-dependent positional distortion. We further counter the spectral bias inherent to implicit neural decoders with a physics-grounded, resolution-gated Fourier decoder that provides anti-aliasing. Pretrained on eight surveys and benchmarked against seven baselines, our model reaches state-of-the-art performance with only 12.2M parameters while outperforming substantially larger models, indicating that encoding physical structure can substitute for parameter scale.
Large language model agents have enabled software development through natural-language interaction, but extending this paradigm to end-to-end game development remains challenging. Building a playable game requires coordinating code, scenes, assets, animations, physics, and configurations, while allowing developers to validate their integrated behavior through interactive testing. We introduce VibeGame, a framework for creating complete 2D game projects from natural-language prompts and optional reference images. Its AI-native game engine represents project state as schema-validated structured text and provides frame-synchronous runtime control, enabling agents to inspect game states, execute actions, and verify behavior at their own pace. Built upon this interface, we design an Adversarial Agent Team of eight specialized agents to separate implementation from review, so that problems are found and fixed through multi-round loops. For each conducted and accepted project, we further introduce a training-free self-evolution mechanism to distill it into reusable skeletons, modules, and collaboration contracts for future development. We evaluate prompt-assets-to-game (PA2G), where agents are given an asset library, and prompt-to-game (P2G), where agents generate their own art. On a GameCraft-Bench subset, VibeGame in P2G consistently outperforms a range of agent harnesses in either P2G or PA2G with the same underlying model, including GPT-6.1-Sol.
Large language models have shown outstanding performance on text-first tasks such as coding, math, and writing. Recent efforts are shifting models' training to include scientific problems that often live in highly specialized domains beyond text, such as protein structure and biological images. However, in these scenarios, their general-purpose accuracy compared with specialized algorithms remains in question. One such domain is cell tracking, which requires spatiotemporal awareness of cell dynamics. To evaluate this, we created a new benchmark, CellTrack-Bench, which assesses Vision-Language Models' (VLMs) cell-tracking-related performance against specialized trackers. The full benchmark draws from 15 standard cell-tracking datasets across four imaging modalities and comprises 1,537 problems corresponding to 9,041 cell-linking decisions. We selected a hard subset requiring greater reasoning effort by identifying failures shared across distinct algorithmic trackers. With this proposed dataset, we: (i) evaluated different environment harnesses for VLMs; (ii) benchmarked 9 frontier closed-source models and 13 open-weight models; and (iii) provided a detailed analysis of their failure modes. Code, the stored model outputs, and a pipeline that regenerates the benchmark from the source datasets are available at https://github.com/royerlab/celltrack-bench.
Recent advances in search agents have enabled large language models to tackle complex tasks by retrieving and synthesizing external information. However, existing search agents often lack search-state awareness, i.e., the ability to recognize what information has been obtained and what remains missing, leading to hallucinated reasoning and prematurely terminated search. To address this limitation, we propose Self-Assessment Reinforcement Learning (SARL), a reinforcement learning framework in which models learn to recognize their own search state during training and improve their search behavior accordingly, without requiring additional assessment at test time. SARL achieves this through a dual reward design that captures the corrective effect of each assessment while penalizing excessive reliance on it, jointly encouraging effective and efficient search trajectories. Under this setup, SARL achieves a 24.3% relative improvement over the strongest baseline, with average gains of 35.9% on multi-hop reasoning and 27.9% on deep research benchmarks against the strongest per-benchmark baselines. SARL further surpasses methods built on auxiliary supervision or multi-agent coordination, despite operating as a standalone RL framework with no inference-time scaffolding. Beyond final performance, SARL maintains training stability under extended search horizons, a regime where existing methods consistently degrade, enabling sustained gains as task complexity grows.
Reinforcement learning (RL) is an effective way to strengthen the long-context reasoning of large language models (LLMs), but it needs training data that are long, diverse, challenging, and verifiable. Human annotation of such data is expensive, naturally occurring data are too scarce to sustain RL, and existing synthesis methods depend on external corpora or stronger teacher models, which become harder to obtain as the trained models get stronger. Recently, RL from zero data, in which a model learns from tasks it proposes itself, has achieved excellent performance, but existing methods target short-context tasks. Therefore, we propose $\textbf{LongZero}$, a method that lets an LLM improve its own long-context reasoning without any external data. Instead of writing questions and answers over existing documents, LongZero first has the model synthesize the question and answer with its supporting evidence in a compact structured form, and only then separately expands the evidence into a long multi-document context. LongZero combines these self-written multi-hop reasoning tasks with self-written multi-round retrieval tasks and trains the model with GRPO. We evaluate LongZero on two LLMs across five long-context reasoning benchmarks and compare it with RL on data synthesized from external corpora. Experimental results show that LongZero can generalize to real tasks, and outperforms both baselines, suggesting a viable and promising path to improving long-context reasoning of LLMs.
A benchmark is not merely a collection of test questions; it naturally induces an empirical measure over difficulty. In this work, we develop this measure-based view of benchmark evaluation. This perspective allows us to identify and formalize two central failure modes: measure mismatch, where the benchmark difficulty distribution differs from the intended difficulty distribution, and coverage gaps, where uncovered regions along the difficulty axis prevent benchmarks from detecting meaningful localized training progress. We propose two diagnostic metrics that capture these two failure modes and provide rigorous theoretical guarantees. Empirically, we analyze 25 popular benchmarks spanning 7 domains using carefully estimated item difficulties. Our results show that many benchmarks implicitly overweight certain difficulty regions, so their leaderboard conclusions should be interpreted with caution; others leave substantial portions of the difficulty axis uncovered, limiting their ability to monitor the model improvement process. Finally, we turn diagnosis into action by introducing BenchFill, a simple difficulty-targeted item rewriting algorithm that effectively reduces local coverage gaps.
Multimodal large language models (MLLMs) show promising reasoning capabilities, but reliable physical interaction requires planning that accounts for when actions are executable and how they change the world. These preconditions and effects are environment-dependent and often not captured by MLLMs, making adaptation challenging. Yet existing approaches either require costly environment-specific training or rely on unstructured interaction experience that is difficult to reuse. Therefore, we introduce CausGround, a non-parametric causal interaction grounding framework that distills interaction experience into reusable action-centered causal knowledge. CausGround identifies action preconditions and effects and organizes them in a factorized world-state representation capturing object attributes and inter-object relations. At planning time, task-relevant causal dependencies are grounded in current observations to support backward reasoning over preconditions and forward reasoning over effects for action selection and multi-step planning. We evaluate CausGround on simulated high-level planning, interaction reasoning, and real-world robot manipulation. On adapted VirtualHome and RoboCasa benchmarks, CausGround improves vision-based high-level planning across MLLM backbones, increasing average task success from 24.9% to 53.3% on VirtualHome and closed-loop planning success from 43.0% to 57.0% on RoboCasa. On SWITCH, average interaction reasoning accuracy increases from 25.8% to 31.9%, indicating improved action selection, recovery, and verification. Moreover, real-world experiments show that CausGround reasons backward from MLLM-interpreted goals to derive executable high-level action sequences for sequential manipulation. These results demonstrate that reusable action preconditions and effects support long-horizon embodied planning across diverse environments without parameter updates.
Simran Khanuja, Rohit Gandikota, Gaurav Parmar, Graham Neubig
Text-to-image (T2I) platforms typically return several images per prompt so users can explore different visual interpretations and choose one that matches their intent. However, T2I models are usually optimized for single-image quality which comes at a cost of multi-image diversity, limiting this exploration. Diversity is also contextual: one user wants variations in background, while another wants them in artistic style, and the same user may later want variations in clothing. The generated images should differ along whatever axis the user cares about. In this paper, we introduce T-DIG, a text-conditioned diversity framework for image generation and editing. T-DIG is an iterative framework comprising multiple modules that lets users type prompts, inspect default generations, request variations using natural language, and continue refining outputs until they're satisfied. At its core, T-DIG uses an LLM policy that produces K axis-conditioned prompts or edit instructions for a frozen generator or editor. We train this policy with GRPO using a text-conditioned diversity reward and alignment penalties. Because the image generator remains frozen, T-DIG applies to open-weight, distilled, and proprietary models. We also release the first dataset for text-conditioned prompt diversity in image generation, with the training and validation splits built from human-written axes and a test set drawn from real user interactions with T-DIG. T-DIG far exceeds baselines in improving text-conditioned diversity, while preserving base-prompt alignment across a variety of metrics and human judgements. We show that user-directed diversity exploration can be added to a wide range of image generation models, open and closed, without the need to access their weights. Website: https://image-study.org/t-dig/
Xuyuan Xiong, Siyu Chen, Zihan Shao, Yue Su, Weiyang Jin, Ping Luo, Zhuoran Yang, Xihui Liu
Looped transformers reuse a shared block of layers to increase computational depth without increasing the number of trainable parameters. They are trained with backpropagation through time (BPTT). To save training compute, BPTT is often truncated to the last few loops, but truncation yields biased gradients that may weaken the benefit of looping. In this work, we study the value of truncation in looped transformer training in a systematic way: we jointly vary model size, training tokens, and forward and backward loop counts across nearly 600 training runs and fit a unified scaling law that captures the effects of looping and truncation through effective parameters. Despite the training-compute savings of truncation, our fitted law favors full BPTT when forward and backward loop counts are jointly optimized. In matched-compute runs, full BPTT through fewer loops consistently outperforms truncated settings of the same compute. The law further shows that at fixed model size, the compute-optimal loop count and training-token budget both grow with compute. When model size is also free, the law identifies two fully trained loops as the best looped setting. Experimentally, this setting achieves lower validation loss than the compute-optimal non-looped model with about 27% fewer non-embedding parameters and only 8% more training compute.
We introduce OvisOCR3, a 0.8B model for a broad range of OCR-related tasks. OvisOCR3 unifies document parsing, general OCR, key information extraction (KIE), document figure captioning, and cross-page table reconstruction in a single model. We train the model on real and synthetic data through continued pretraining and use domain-balanced mid-training to balance supervision across tasks with different data volumes and response lengths. Multi-teacher on-policy distillation transfers expertise from five domain-specific 4B teachers into the shared model, followed by mixed-task reinforcement learning with domain-specific rewards. We compare OvisOCR3 with general-purpose vision-language models and specialized OCR models on public and in-house benchmarks covering all five task families. In these comparisons, it achieves an overall score of 96.74 on OmniDocBench v1.6 and the highest overall scores on PureDocBench, Real5-OmniDocBench, and MDPBench. On OCRBench v2, it ranks first, scoring 65.9 and 61.3 on the English and Chinese splits, respectively. It also outperforms the baselines on OCRBench and all four KIE benchmarks. On DocFigCapBench, it achieves the highest localization F1 and a caption score close to that of Qwen3.8-Max. For cross-page tables, it leads in Relation F1 on CrossTabBench and truncated-table TEDS on MPDocBench-Parse. These results demonstrate that a compact 0.8B model can achieve strong performance across diverse OCR-related tasks. OvisOCR3 is available at https://huggingface.co/ATH-MaaS/OvisOCR3.
In long-term interactions with large language model (LLM) agents, users reveal habits through their choices and actions. These habits capture implicit user information that agents must infer across conversations. However, individual observations may provide only partial evidence about a habit or when it applies. Retaining past interactions alone does not ensure that agents can infer habits and apply them appropriately. We introduce HABIT-Bench to evaluate how well agent memory systems infer habits from weak evidence and apply them to later requests. We construct synthetic user histories from task-oriented dialogue data and controlled habit graphs. The histories span hundreds of sessions across food, finance, software, and travel. Four-choice tasks assess agents' ability to infer user habits and apply them appropriately. They include temporary exceptions, changes in habits, and suggestions that users have not accepted. Sampled tasks undergo independent human review. We evaluate eight memory methods with Qwen3-235B-A22B and GLM-5.3-Flash. The best-performing memory method achieves 33.61\% macro accuracy across the four domains. Performance remains limited even when annotated evidence is supplied directly. These results expose a persistent gap between access to past interactions and the ability to infer and appropriately use user habits.
George Nigmatulin, Valentin Mohl, Sascha Frey, Kang Li, Aramis Marti-Shahandeh, Mihai Cucuringu, Anisoara Calinescu, Jakob Foerster 等 9 位作者
Realistic market impact is essential for generative limit order book (LOB) models to serve as world models for trading agents. If a simulator misprices this mechanism, agents learn policies that are optimal in the simulator but wrong in the real market. Existing metrics assess generated streams purely through distributional comparison with historical data, leaving counterfactual response to an agent's own orders largely unexplored. Here, we show that market impact emerges naturally in neural LOB generators trained solely for next-token prediction. We inject execution-style metaorders into the generated message flow of five autoregressive models across five NASDAQ stocks and measure the response using estimators from the empirical impact literature, validated on baselines whose physics is known in advance. Our benchmark evaluates any message-level generator in three phases: free generation (reading order-flow persistence against price diffusivity); metaorder injection (yielding the impact law's exponent and amplitude); and a post-execution window (measuring relaxation). In addition, comparing each visible injection to a blind, simulator-only control isolates the model's learned response from standard book mechanics. For every model on every stock, we measure how the impact responds as a function of cumulative metaorder participation. Interestingly, we observe emergent impact in generative models of market data, with exponents in the range of 0.63 to 1.21, slightly above the traditional 0.5 of the square-root law of impact. In the family of models tested, we are so far unable to observe impact decay during relaxation. We tentatively attribute this to the still-short context windows of the models. We release the benchmark as open source to validate future generative models for market data.
Synthesizing a monocular video from a new camera trajectory requires preserving the observed scene geometry and motion while completing regions that become newly visible. Feed-forward 4D reconstructors recover the structure of the observed scene, whereas video diffusion models provide a generative prior for completing unseen content. We introduce Holo4D, a video reshooting framework that uses reconstructed 3D trajectories as a holistic source-view 4D scaffold. We project these trajectories into the requested views and encode their geometry and source appearance to guide a pretrained video diffusion model. Together with the source video and target cameras, the scaffold grounds generation in reconstructed observations while leaving unsupported regions to the generative prior. Experiments demonstrate the value of trajectory-based conditioning for camera control and novel-view synthesis. The framework accommodates different 4D reconstructors without retraining the video generator, allowing it to benefit from future advances in source reconstruction. More details can be found in our project page.
Blockchain forensics is adversarial and open-ended: money launderers split, swap, and bridge funds, cut their trail at mixers or custodial services, and change tactics from one incident to the next. Investigators must keep adapting their tracing in response. Yet the immutability of public ledgers preserves every recorded movement in public view, making much of the investigation independently verifiable. We introduce ABF-Bench, an interactive benchmark that makes evidence-backed fund tracing a verifiable agent task, in which every claimed movement must cite ledger records the agent retrieved. A case-to-task pipeline builds well-posed tasks across ledgers: conservative rules halt tracing wherever the records leave a continuation unresolved, and each task's reference graph is the answer those rules determine. Built from real laundering activity, the tasks combine multi-hop transfers, DeFi exchanges, cross-chain round trips, custodial and mixer deposits, and address poisoning. We also propose the ABF agent harness, built around a working memory, and evaluate it alongside general-purpose agents and fixed-order tracers. The harness reaches 75% edge recall, against 47% for Codex with the same model and 10% for Claude Code, which loses recall to overconfidence: it cites records it left unretrieved. Compared with fixed-order tracers, the harness acquires evidence faster than the best of them early in the budget and recognizes undisclosed custodial services that they walk past (89% vs. 0%), with 98% fidelity. However, agents tend to stop early and lose recall on larger tasks, which holds their final recall below the best tracer. Beyond exposing these gaps, ABF-Bench opens blockchain forensics to LLM agents as a verifiable domain, where the verdicts that score agents also offer potential training signals.
Computer-use agents (CUAs) are commonly evaluated against fixed task specifications, but users may omit requirements, add constraints, or require approval before consequential actions. Completing the benchmark instruction does not guarantee serving the person who asked. We introduce PersonaCUABench, which holds 74 tasks across 11 domains fixed and varies only who asks. A persona agent, playing one of 44 personas in four groups, rewrites each instruction in that user's own words and answers the CUA's questions, while the CUA sees only the dialogue, never the profile or the original instruction. Across nine backbones (3,330 runs), mean task completion falls by 36% relative to the original requests. The losses have two sources: in 59.9% of cells that lose points, the agent engages with what the user asked for and still gets it wrong, and 11.2% are stops the user requested that the checker scores as failures. Even when a persona request is fully completed, the user gives the top satisfaction rating in only 62.7% of cells, against 84.1% for the original requests. CUAs should therefore be evaluated not only on task completion but also on whether they meet the requirements and constraints users express. Code: https://github.com/MatrAIx-ai/PersonaCUABench. Project page: https://matraix.ai/research/persona-cua.html
Forecasting an unresolved event requires finding relevant evidence and deciding how strongly it supports each outcome. Yet outcome-based training of Large Language Model (LLM) forecasters often begins with preassembled evidence at a single prediction date, leaving evidence acquisition outside the learning loop. We introduce AlphaBet, an agentic reinforcement learning (RL) framework that trains a single LLM policy to search, reason, and report probabilities across multiple pre-resolution dates. Each dated forecast is scored by a proper scoring rule against the eventual outcome, allowing one resolved event to supervise decisions under different information states. We decompose forecasting regret into an evidence-acquisition gap and a probability-reporting gap, and show that Group Relative Policy Optimization (GRPO)'s standard-deviation normalization can distort the reporting target under multi-class proper scores. On GLM-4.5-Air, AlphaBet reduces the Brier score from 0.731 to 0.626 on Forecast-Dojo and from 0.612 to 0.528 on FutureX. Controlled ablations attribute these gains to multi-date training and show that they come from improvements in both information acquisition and probability reporting. Our empirical results validate that GRPO's std normalization degrades proper-scoring-rule performance while leaving top-outcome accuracy largely unchanged.
Yunqi Liu, Peiyu Wang, Bingli Wang, Leo Luo, Xuanwu Song, Xudong Ma, Hongxin Li, Yaxin Du 等 24 位作者
Command-line interface (CLI) agents increasingly execute longer-horizon tasks, yet existing benchmarks either assume that user instructions remain fixed throughout execution or evaluate task outcomes without capturing runtime collaboration quality. In this paper, we present REALCLI, a paired benchmark that unifies the evaluation of task execution and human-agent collaboration quality on realistic CLI tasks. To comprehensively measure collaboration, REALCLI builds a five-dimensional framework, operationalized into 22 sub-dimensions and 51 interaction scenario templates. REALCLI provides two modes for paired comparison: REALCLI-EXEC supplies complete specifications upfront, whereas REALCLI-COLLAB introduces a scripted user that delivers progressive information, corrections, constraint revisions, and authorization decisions. Evaluations reveal profound capability gaps in frontier agents: the best configuration attains an average score of only 49.2%, and when evaluated by the joint collaboration score ($J_{\mathrm{collab}}$), the peak performance reaches merely 37.6%. These results show that task effectiveness diverges sharply from collaboration quality, with models consistently struggling in steerability, trust calibration, and safety boundary management. REALCLI provides a rigorous, reproducible foundation for measuring and advancing collaborative CLI agents.
Sparse autoencoders (SAEs) decompose language model activations into sparse features, but prompt transformations designed to preserve the underlying request can redistribute activity across them. A feature change observed on one transformed prompt need not recur, and even a recurring change need not affect a downstream computation. We introduce the redistribution sensitivity score (RSS), which identifies large feature responses that recur in a consistent direction, and evaluate the resulting supports on variants excluded from selection. Across 27 pretrained SAEs spanning eight language models, the tested prompt transformations produce a reproducible redistribution profile on these excluded variants. We then separate recurring representation change from signed readout contribution and decision failure. For a fixed linear readout, an exact decomposition shows how response magnitude, readout coefficients, directional alignment, and the clean margin determine downstream consequences. In two frozen safety readouts, combining RSS support erosion with the full clean margin yields recurrent failure AUROCs of 0.935 and 0.958 using median supports of 20 and 19 features, compared with 0.965 and 0.964 using full erosion. RSS therefore provides a way to identify recurring sparse representation changes and quantify their contribution to a downstream readout without equating representation change with failure. Code is available at https://github.com/OpenSAE/Redistribution-Sensitivity-Score
Yingzhuo Deng, Chong Li, Bingxin Liu, Yi Shu, Tianyu Peng, Can Xie, Siyao Song, Shuo Ren 等 9 位作者
Spoken Dialogue Models (SDMs) typically share a speech-token space for generation and synthesis, leading to long autoregressive sequences and costly inference. Lower-rate speech tokenizers shorten sequences but may introduce compression-quality trade-offs and costly replacement without breaking this coupling. To address this limitation, we propose CoSTA, a compatibility-preserving post-hoc framework that decouples generation and synthesis token spaces while retaining the original speech tokenizer and decoder. CoSTA extends the generation token space with compressed representations via mean-initialized vocabulary extension, pronunciation adaptation, and dialogue adaptation, followed by deterministic expansion back to the original speech-token space before synthesis. On bilingual speech-dialogue benchmarks, CoSTA achieves approximately $1.7\times$ expanded speech-token generation speed, lower full-response latency, and higher throughput while preserving response quality. Analyses show more sustained interaction under limited budgets and a practical trade-off among compression, decoding efficiency, and quality. Further experiments demonstrate cross-architecture transfer and complementarity to low-rate tokenization.
Anita Kriz, Antonin Berthon, Nicolás Astorga, Mihaela van der Schaar
Despite their extensive pretraining knowledge and strong reasoning capabilities, large language models (LLMs) struggle to adaptively gather information in experimental design tasks. We study how to amortize sequential Bayesian experimental design (BED) into an LLM by explicitly training the reasoning that guides action selection using information-gain rewards. Our approach specifies the experimental generative model independently of the LLM's own beliefs, allowing us to optimize expected information gain under a model of our choosing. We investigate the design choices for amortization, including alternative generative-model constructions, reward estimation, hypothesis-space specification, multi-step objectives, and computational costs relative to training- and inference-time BED baselines. On Twenty Questions, the resulting policies improve information acquisition and task performance, outperform inference- and training-time BED baselines at substantially lower computational cost, and generalize to unseen categories. We further demonstrate positive transfer from Twenty Questions training to Medi-Q, a real-world clinical dialogue benchmark. These results suggest that information gathering can be learned as a generalizable reasoning skill in LLMs.
Modern image generation largely relies on specialized generative models that leave the process of image construction implicit in learned parameters. We explore an alternative: general-purpose coding agents that construct images through executable programs. Our study of zero-shot drawing on AgentDrawBench, a 200-prompt benchmark, reveals a distinctive capability profile. Agents excel at preserving structure and satisfying precise visual constraints, achieving competitive results in domains where compositional accuracy matters more than photorealistic appearance, while remaining limited in perceptual richness. Their visual expressiveness varies with the drawing interface, and a qualitative study on reconstruction shows how procedural skills extracted from human demonstrations support different artistic media. We then examine code as an external visual language of thought: a structured, executable representation that agents can construct and inspect. We test it in visual question answering task and find reconstructed visual code retains substantial information for question answering and, in some domains, supports higher accuracy than the original image. These abilities raise a further question: do agents draw on a shared internal representation of scene content beyond a particular drawing program? We explore an internal visual language-of-thought hypothesis, in which a shared, compositionally structured representation supports visual operations across drawing languages. Cross-harness drawings and images assembled from independent single-pixel queries exhibit related layouts and tones. Agents also maintain self-generated room arrangements across viewpoints after visible drawings are removed. These results demonstrate that the coding agent would maintain a consistent and coherent world inside its imagine even before the harness gets involved and the generation starts.
Many Indian banks serve customers through FAQ pages and scripted chatbots. Assistants based on language models could respond faster and carry out requests such as blocking a card, but only if they give correct information and follow the bank's rules. These rules prescribe a procedure, a set of required steps, such as obtaining a closure quote before closing a deposit. Adapting models to this domain requires measuring how well they follow the procedure, before and after training. Benchmarks usually judge only the records left at the end of a conversation. Many required steps, however, leave the records unchanged, so an assistant that skips them can still be judged successful. We present TauIndianBankBench, an action-oriented benchmark of 1,000 Indian retail-banking tasks where an assistant helps a simulated customer. Besides the final records, it checks that the assistant took every required step and made no unrequested change. For every model, 6.5-12.0% of sessions are judged successful when only the final records are checked, although the assistant skipped a required step or made an unrequested change. Scoring both procedure and outcome shows more accurately how well language models perform in Indian retail banking.