Searcharxiv⌕ Search

arXiv subjects

Tuhin Kundu

Publications and source records attributed to Tuhin Kundu.

6 recordsLinked to original sources

A Hierarchical Statistical Framework for Evaluating Agent Reliability

Agents operating in complex scenarios often face uncertainty due to incomplete information about the environment they operate in, ambiguous user queries, complex business rules and the inherently stochastic dynamics of the governing policy. Several new benchmarks have been proposed to evaluate LLM-based agents on complex tasks in multi-tool environments, including settings with policy constraints and varying levels of intent ambiguity. These conditions make it important to evaluate agentic systems systematically and quantify their performance under different conditions to assess their reliability in production. This need is evident in the recently introduced Thinkingbox benchmark, where even the strongest models struggle to perform consistently across diverse business scenarios. This paper presents a statistical framework for evaluating uncertainty and reliability in agentic systems. We define a hierarchy of scenarios, instances, and samples to assess model performance and variability systematically. We apply this framework to common metrics found in the literature: mean pass rate, $pass@k$, and $passhat^k$, with uncertainty estimates for each metric. Aggregating metrics at different levels of the hierarchy reveals different structures and properties of the policy model; we connect this hierarchical aggregation with the known process of \emph{coarse-graining} as intended in the physics and statistical literature. We independently derive exact finite-sample expressions for the sampling variances of the unbiased $pass@k$ and $passhatk$ estimators; related results also appear in recent submissions. Our central contribution is a constructive derivation of hierarchical uncertainty propagation, making explicit its correspondence with coarse-graining and separating execution noise from task heterogeneity.

cs.AI↗

One Success Isn't Reliability: Thinkingbox, a Sandbox and Benchmark for Agents in Stateful Business Workflows

Recent agent benchmarks increasingly ground evaluation in executable environments, from code repair to web navigation, app APIs, and function calling. Yet completing consequential work beyond code requires more than producing a plausible response or valid tool call: agents must gather missing information over multiple turns, follow domain policies, coordinate dependent tools, and realize the correct persistent state transition without collateral effects. In this paper, we introduce Thinkingbox, a sandbox for tool-agent-user interaction that provides isolated MCP-compatible tool sessions, complete execution traces, and outcome evaluation over terminal backend state. Built on this sandbox, Thinkingbox-bench contains 507 policy-conditioned workflows across business scenarios, including retail, hospitality, auto insurance, neobank internal IT, and consulting IT/HR support. Each attempt is evaluated by task-specific executable checks that accept valid trajectories while rejecting wrong, missing, or extra effects; designated tasks additionally check required properties of the final response. Our experiments reveal that even the strongest proprietary and open-weight models show steep reliability drops: Claude Opus 5 falls from 66.50% pass@1 to 47.53% pass^20, and Kimi-K3 from 57.37% pass@1 to 17.60% pass^20. Moreover, many failed trials terminate cleanly after valid state-changing actions, so response- or tool-call-level signals poorly proxy end-to-end completion. Thinkingbox-bench reveals a large gap between occasionally finding a successful trajectory and reliably completing stateful business tasks. We release Thinkingbox, Thinkingbox-Bench, and Thinkingbox-integrated reinforcement learning framework.

cs.CL↗

Imaginative Perception Tokens Enhance Spatial Reasoning in Multimodal Language Models

Vision language models (VLMs) excel at many tasks but still struggle with spatial reasoning when critical information is not directly observable. Many such problems require imaginative perception: inferring what would be seen from an unseen viewpoint, tracing paths through occluded spaces, or integrating partial observations into a coherent spatial representation. We introduce Imaginative Perception Tokens (IPT), intermediate perceptual representations that externalize what a VLM would perceive under alternative spatial configurations while remaining consistent with the observed input. To study this capability, we formulate three tasks, Perspective Taking (PET), Path Tracing (PT), and Multiview Counting (MVC), and construct datasets of approximately 20K examples with ground truth imaginations, answers, and evaluation benchmarks. Using the unified VLM BAGEL as the backbone, IPT supervision consistently improves spatial reasoning and often outperforms textual chain of thought training, even without generating images at inference time. On MVC, IPT improves accuracy by 3.4% and achieves competitive performance with strong closed-source models on PT. We further find that combining IPT and label-only supervision yields additional gains, whereas textual chain of thought can substantially degrade performance, suggesting a modality mismatch when spatial computation is forced through language. Overall, IPT provides a principled supervision signal for reasoning about unobserved spatial structure, improving generalization while producing interpretable intermediate representations.

cs.AI↗

MIMIC: Masked Image Modeling with Image Correspondences

Dense pixel-specific representation learning at scale has been bottlenecked due to the unavailability of large-scale multi-view datasets. Current methods for building effective pretraining datasets heavily rely on annotated 3D meshes, point clouds, and camera parameters from simulated environments, preventing them from building datasets from real-world data sources where such metadata is lacking. We propose a pretraining dataset-curation approach that does not require any additional annotations. Our method allows us to generate multi-view datasets from both real-world videos and simulated environments at scale. Specifically, we experiment with two scales: MIMIC-1M with 1.3M and MIMIC-3M with 3.1M multi-view image pairs. We train multiple models with different masked image modeling objectives to showcase the following findings: Representations trained on our automatically generated MIMIC-3M outperform those learned from expensive crowdsourced datasets (ImageNet-1K) and those learned from synthetic environments (MULTIVIEW-HABITAT) on two dense geometric tasks: depth estimation on NYUv2 (1.7%), and surface normals estimation on Taskonomy (2.05%). For dense tasks which also require object understanding, we outperform MULTIVIEW-HABITAT, on semantic segmentation on ADE20K (3.89%), pose estimation on MSCOCO (9.4%), and reduce the gap with models pre-trained on the object-centric expensive ImageNet-1K. We outperform even when the representations are frozen, and when downstream training data is limited to few-shot. Larger dataset (MIMIC-3M) significantly improves performance, which is promising since our curation method can arbitrarily scale to produce even larger datasets. MIMIC code, dataset, and pretrained models are open-sourced at https://github.com/RAIVNLab/MIMIC.

cs.CV↗

Neural Keyphrase Generation: Analysis and Evaluation

Keyphrase generation aims at generating topical phrases from a given text either by copying from the original text (present keyphrases) or by producing new keyphrases (absent keyphrases) that capture the semantic meaning of the text. Encoder-decoder models are most widely used for this task because of their capabilities for absent keyphrase generation. However, there has been little to no analysis on the performance and behavior of such models for keyphrase generation. In this paper, we study various tendencies exhibited by three strong models: T5 (based on a pre-trained transformer), CatSeq-Transformer (a non-pretrained Transformer), and ExHiRD (based on a recurrent neural network). We analyze prediction confidence scores, model calibration, and the effect of token position on keyphrases generation. Moreover, we motivate and propose a novel metric framework, SoftKeyScore, to evaluate the similarity between two sets of keyphrases by using softscores to account for partial matching and semantic similarity. We find that SoftKeyScore is more suitable than the standard F1 metric for evaluating two sets of given keyphrases.

cs.CL↗

KPDrop: Improving Absent Keyphrase Generation

Keyphrase generation is the task of generating phrases (keyphrases) that summarize the main topics of a given document. Keyphrases can be either present or absent from the given document. While the extraction of present keyphrases has received much attention in the past, only recently a stronger focus has been placed on the generation of absent keyphrases. However, generating absent keyphrases is challenging; even the best methods show only a modest degree of success. In this paper, we propose a model-agnostic approach called keyphrase dropout (or KPDrop) to improve absent keyphrase generation. In this approach, we randomly drop present keyphrases from the document and turn them into artificial absent keyphrases during training. We test our approach extensively and show that it consistently improves the absent performance of strong baselines in both supervised and resource-constrained semi-supervised settings.

cs.CL↗