arXiv subjects

Pavan Chakraborty

Publications and source records attributed to Pavan Chakraborty.

At least 19 recordsLinked to original sources

arXiv2026

Measuring and curing reasoning rigidity: from decorative chain-of-thought to genuine faithfulness

Language models increasingly show their work by writing step-by-step reasoning before answering. But are these steps genuinely used, or is the answer rigid - fixed before reasoning begins? We introduce the Step-Level Reasoning Capacity (SLRC) metric and prove it is a consistent causal estimator (Theorem 1). We propose LC-CoSR, a training method with Lyapunov stability guarantees that directly reduces rigidity. Evaluating 16 frontier models (o4-mini, GPT-5.4, Claude Opus, Grok-4, DeepSeek-R1, Gemini 2.5 Pro, and others) across six domains at N=133-500, we find reasoning falls into three modes. OpenAI's o4-mini shows 74-88% step necessity on five of six tasks (73.8-88.3%) - the highest SLRC in our study. The critical differentiator is RL-based reasoning training, not thinking tokens: Grok-4's reasoning mode shows lower faithfulness than its non-reasoning mode (1.4% vs 7.2% necessity). We discover a faithfulness paradox - high-SLRC models are more susceptible to sycophancy - and propose the Reasoning Integrity Score (RIS = SLRC x (1-Sycophancy)), which significantly predicts error detection (rho=0.66, p=0.026). LC-CoSR achieves 2.6x less negative reward than FARL and CSR baselines without external model dependencies.

cs.CL↗

arXiv2026

When Names Change Verdicts: Intervention Consistency Reveals Systematic Bias in LLM Decision-Making

Large language models (LLMs) are increasingly used for high-stakes decisions, yet their susceptibility to spurious features remains poorly characterized. We introduce ICE-Guard, a framework applying intervention consistency testing to detect three types of spurious feature reliance: demographic (name/race swaps), authority (credential/prestige swaps), and framing (positive/negative restatements). Across 3,000 vignettes spanning 10 high-stakes domains, we evaluate 11 LLMs from 8 families and find that (1) authority bias (mean 5.8%) and framing bias (5.0%) substantially exceed demographic bias (2.2%), challenging the field's narrow focus on demographics; (2) bias concentrates in specific domains -- finance shows 22.6% authority bias while criminal justice shows only 2.8%; (3) structured decomposition, where the LLM extracts features and a deterministic rubric decides, reduces flip rates by up to 100% (median 49% across 9 models). We demonstrate an ICE-guided detect-diagnose-mitigate-verify loop achieving cumulative 78% bias reduction via iterative prompt patching. Validation against real COMPAS recidivism data shows COMPAS-derived flip rates exceed pooled synthetic rates, suggesting our benchmark provides a conservative estimate of real-world bias. Code and data are publicly available.

cs.CL↗

arXiv2026

ICE: Intervention-Consistent Explanation Evaluation with Statistical Grounding for LLMs

Evaluating whether explanations faithfully reflect a model's reasoning remains an open problem. Existing benchmarks use single interventions without statistical testing, making it impossible to distinguish genuine faithfulness from chance-level performance. We show that faithfulness is not a fixed property but an operator-dependent quantity that changes with the intervention method used to measure it. We introduce ICE (Intervention-Consistent Explanation), a framework that evaluates explanations against random baselines of equal size under multiple operators. Evaluating 7 LLMs across 4 tasks with deletion and retrieval infill operators, we find that switching operators crosses the positive-evidence threshold in 18% of configurations (5 of 28 attention comparisons), with gaps reaching 44 percentage points. Randomized baselines detect anti-faithfulness (explanations worse than random) in nearly one-third of English deletion configurations, invisible without random comparisons. These patterns persist across 6 non-English languages and 2 attribution methods. The methodology generalizes to step-level chain-of-thought evaluation, where preliminary results on 3 frontier models suggest that high accuracy does not imply faithful reasoning.

cs.CL↗

arXiv2026

Proof-Carrying Materials: Falsifiable Safety Certificates for Machine-Learned Interatomic Potentials

Machine-learned interatomic potentials (MLIPs) are deployed for high-throughput materials screening without formal reliability guarantees. We show that a single MLIP used as a stability filter misses 93% of density functional theory (DFT)-stable materials (recall 0.07) on a 25,000-material benchmark. Proof-Carrying Materials (PCM) closes this gap through three stages: adversarial falsification across compositional space, bootstrap envelope refinement with 95% confidence intervals, and Lean 4 formal certification. Auditing CHGNet, TensorNet and MACE reveals architecture-specific blind spots with near-zero pairwise error correlations (r <= 0.13; n = 5,000), confirmed by independent Quantum ESPRESSO validation (20/20 converged; median DFT/CHGNet force ratio 12x). A risk model trained on PCM-discovered features predicts failures on unseen materials (AUC-ROC = 0.938 +/- 0.004) and transfers across architectures (cross-MLIP AUC-ROC ~ 0.70; feature importance r = 0.877). In a thermoelectric screening case study, PCM-audited protocols discover 62 additional stable materials missed by single-MLIP screening - a 25% improvement in discovery yield.

cond-mat.mtrl-sci↗

arXiv2026

Budget-Sensitive Discovery Scoring: A Formally Verified Framework for Evaluating AI-Guided Scientific Selection

Scientific discovery increasingly relies on AI systems to select candidates for expensive experimental validation, yet no principled, budget-aware evaluation framework exists for comparing selection strategies -- a gap intensified by large language models (LLMs), which generate plausible scientific proposals without reliable downstream evaluation. We introduce the Budget-Sensitive Discovery Score (BSDS), a formally verified metric -- 20 theorems machine-checked by the Lean 4 proof assistant -- that jointly penalizes false discoveries (lambda-weighted FDR) and excessive abstention (gamma-weighted coverage gap) at each budget level. Its budget-averaged form, the Discovery Quality Score (DQS), provides a single summary statistic that no proposer can inflate by performing well at a cherry-picked budget. As a case study, we apply BSDS/DQS to: do LLMs add marginal value to an existing ML pipeline for drug discovery candidate selection? We evaluate 39 proposers -- 11 mechanistic variants, 14 zero-shot LLM configurations, and 14 few-shot LLM configurations -- using SMILES representations on MoleculeNet HIV (41,127 compounds, 3.5% active, 1,000 bootstrap replicates) under both random and scaffold splits. Three findings emerge. First, the simple RF-based Greedy-ML proposer achieves the best DQS (-0.046), outperforming all MLP variants and LLM configurations. Second, no LLM surpasses the Greedy-ML baseline under zero-shot or few-shot evaluation on HIV or Tox21, establishing that LLMs provide no marginal value over an existing trained classifier. Third, the proposer hierarchy generalizes across five MoleculeNet benchmarks spanning 0.18%-46.2% prevalence, a non-drug AV safety domain, and a 9x7 grid of penalty parameters (tau >= 0.636, mean tau = 0.863). The framework applies to any setting where candidates are selected under budget constraints and asymmetric error costs.

cs.LG↗

arXiv2026

Contextual StereoSet: Stress-Testing Bias Alignment Robustness in Large Language Models

A model that avoids stereotypes in a lab benchmark may not avoid them in deployment. We show that measured bias shifts dramatically when prompts mention different places, times, or audiences -- no adversarial prompting required. We introduce Contextual StereoSet, a benchmark that holds stereotype content fixed while systematically varying contextual framing. Testing 13 models across two protocols, we find striking patterns: anchoring to 1990 (vs. 2030) raises stereotype selection in all models tested on this contrast (p<0.05); gossip framing raises it in 5 of 6 full-grid models; out-group observer framing shifts it by up to 13 percentage points. These effects replicate in hiring, lending, and help-seeking vignettes. We propose Context Sensitivity Fingerprints (CSF): a compact profile of per-dimension dispersion and paired contrasts with bootstrap CIs and FDR correction. Two evaluation tracks support different use cases -- a 360-context diagnostic grid for deep analysis and a budgeted protocol covering 4,229 items for production screening. The implication is methodological: bias scores from fixed-condition tests may not generalize.This is not a claim about ground-truth bias rates; it is a stress test of evaluation robustness. CSF forces evaluators to ask, "Under what conditions does bias appear?" rather than "Is this model biased?" We release our benchmark, code, and results.

cs.CL↗

arXiv2025

Polyp detection in colonoscopy images using YOLOv11

Colorectal cancer (CRC) is one of the most commonly diagnosed cancers all over the world. It starts as a polyp in the inner lining of the colon. To prevent CRC, early polyp detection is required. Colonosopy is used for the inspection of the colon. Generally, the images taken by the camera placed at the tip of the endoscope are analyzed by the experts manually. Various traditional machine learning models have been used with the rise of machine learning. Recently, deep learning models have shown more effectiveness in polyp detection due to their superiority in generalizing and learning small features. These deep learning models for object detection can be segregated into two different types: single-stage and two-stage. Generally, two stage models have higher accuracy than single stage ones but the single stage models have low inference time. Hence, single stage models are easy to use for quick object detection. YOLO is one of the singlestage models used successfully for polyp detection. It has drawn the attention of researchers because of its lower inference time. The researchers have used Different versions of YOLO so far, and with each newer version, the accuracy of the model is increasing. This paper aims to see the effectiveness of the recently released YOLOv11 to detect polyp. We analyzed the performance for all five models of YOLOv11 (YOLO11n, YOLO11s, YOLO11m, YOLO11l, YOLO11x) with Kvasir dataset for the training and testing. Two different versions of the dataset were used. The first consisted of the original dataset, and the other was created using augmentation techniques. The performance of all the models with these two versions of the dataset have been analysed.

cs.CV↗

arXiv2024

Point-GR: Graph Residual Point Cloud Network for 3D Object Classification and Segmentation

In recent years, the challenge of 3D shape analysis within point cloud data has gathered significant attention in computer vision. Addressing the complexities of effective 3D information representation and meaningful feature extraction for classification tasks remains crucial. This paper presents Point-GR, a novel deep learning architecture designed explicitly to transform unordered raw point clouds into higher dimensions while preserving local geometric features. It introduces residual-based learning within the network to mitigate the point permutation issues in point cloud data. The proposed Point-GR network significantly reduced the number of network parameters in Classification and Part-Segmentation compared to baseline graph-based networks. Notably, the Point-GR model achieves a state-of-the-art scene segmentation mean IoU of 73.47% on the S3DIS benchmark dataset, showcasing its effectiveness. Furthermore, the model shows competitive results in Classification and Part-Segmentation tasks.

cs.CV↗

arXiv2024

Inverse kinematics learning of a continuum manipulator using limited real time data

Data driven control of a continuum manipulator requires a lot of data for training but generating sufficient amount of real time data is not cost efficient. Random actuation of the manipulator can also be unsafe sometimes. Meta learning has been used successfully to adapt to a new environment. Hence, this paper tries to solve the above mentioned problem using meta learning. We consider two cases for that. First, this paper proposes a method to use simulation data for training the model using MAML(Model-Agnostic Meta-Learning). Then, it adapts to the real world using gradient steps. Secondly,if the simulation model is not available or difficult to formulate, then we propose a CGAN(Conditional Generative adversial network)-MAML based method for it. The model is trained using a small amount of real time data and augmented data for different loading conditions. Then, adaptation is done in the real environment. It has been found out from the experiments that the relative positioning error for both the cases are below 3%. The proposed models are experimentally verified on a real continuum manipulator.

cs.RO↗

arXiv2024

Hybrid CNN Bi-LSTM neural network for Hyperspectral image classification

Hyper spectral images have drawn the attention of the researchers for its complexity to classify. It has nonlinear relation between the materials and the spectral information provided by the HSI image. Deep learning methods have shown superiority in learning this nonlinearity in comparison to traditional machine learning methods. Use of 3-D CNN along with 2-D CNN have shown great success for learning spatial and spectral features. However, it uses comparatively large number of parameters. Moreover, it is not effective to learn inter layer information. Hence, this paper proposes a neural network combining 3-D CNN, 2-D CNN and Bi-LSTM. The performance of this model has been tested on Indian Pines(IP) University of Pavia(PU) and Salinas Scene(SA) data sets. The results are compared with the state of-the-art deep learning-based models. This model performed better in all three datasets. It could achieve 99.83, 99.98 and 100 percent accuracy using only 30 percent trainable parameters of the state-of-art model in IP, PU and SA datasets respectively.

eess.IV↗

arXiv2023

Guidance system for Visually Impaired Persons using Deep Learning and Optical flow

Visually impaired persons find it difficult to know about their surroundings while walking on a road. Walking sticks used by them can only give them information about the obstacles in the stick's proximity. Moreover, it is mostly effective in static or very slow-paced environments. Hence, this paper introduces a method to guide them in a busy street. To create such a system it is very important to know about the approaching object and its direction of approach. To achieve this objective we created a method in which the image frame received from the video is divided into three parts i.e. center, left, and right to know the direction of approach of the approaching object. Object detection is done using YOLOv3. Lucas Kanade's optical flow estimation method is used for the optical flow estimation and Depth-net is used for depth estimation. Using the depth information, object motion trajectory, and object category information, the model provides necessary information/warning to the person. This model has been tested in the real world to show its effectiveness.

cs.CV↗

arXiv2022

Local Gradient Hexa Pattern: A Descriptor for Face Recognition and Retrieval

Local descriptors used in face recognition are robust in a sense that these descriptors perform well in varying pose, illumination and lighting conditions. Accuracy of these descriptors depends on the precision of mapping the relationship that exists in the local neighborhood of a facial image into microstructures. In this paper a local gradient hexa pattern (LGHP) is proposed that identifies the relationship amongst the reference pixel and its neighboring pixels at different distances across different derivative directions. Discriminative information exists in the local neighborhood as well as in different derivative directions. Proposed descriptor effectively transforms these relationships into binary micropatterns discriminating interclass facial images with optimal precision. Recognition and retrieval performance of the proposed descriptor has been compared with state-of-the-art descriptors namely LDP and LVP over the most challenging and benchmark facial image databases, i.e. Cropped Extended Yale-B, CMU-PIE, color-FERET, and LFW. The proposed descriptor has better recognition as well as retrieval rates compared to state-of-the-art descriptors.

cs.CV↗

arXiv2022

Local Quadruple Pattern: A Novel Descriptor for Facial Image Recognition and Retrieval

In this paper a novel hand crafted local quadruple pattern (LQPAT) is proposed for facial image recognition and retrieval. Most of the existing hand-crafted descriptors encodes only a limited number of pixels in the local neighbourhood. Under unconstrained environment the performance of these descriptors tends to degrade drastically. The major problem in increasing the local neighbourhood is that, it also increases the feature length of the descriptor. The proposed descriptor try to overcome these problems by defining an efficient encoding structure with optimal feature length. The proposed descriptor encodes relations amongst the neighbours in quadruple space. Two micro patterns are computed from the local relationships to form the descriptor. The retrieval and recognition accuracies of the proposed descriptor has been compared with state of the art hand crafted descriptors on bench mark databases namely; Caltech-face, LFW, Colour-FERET, and CASIA-face-v5. Result analysis shows that the proposed descriptor performs well under uncontrolled variations in pose, illumination, background and expressions.

cs.CV↗

arXiv2022

Local Directional Gradient Pattern: A Local Descriptor for Face Recognition

In this paper a local pattern descriptor in high order derivative space is proposed for face recognition. The proposed local directional gradient pattern (LDGP) is a 1D local micropattern computed by encoding the relationships between the higher order derivatives of the reference pixel in four distinct directions. The proposed descriptor identifies the relationship between the high order derivatives of the referenced pixel in four different directions to compute the micropattern which corresponds to the local feature. Proposed descriptor considerably reduces the length of the micropattern which consequently reduces the extraction time and matching time while maintaining the recognition rate. Results of the extensive experiments conducted on benchmark databases AT&T, Extended Yale B and CMU-PIE show that the proposed descriptor significantly reduces the extraction as well as matching time while the recognition rate is almost similar to the existing state of the art methods.

cs.CV↗

arXiv2022

R-Theta Local Neighborhood Pattern for Unconstrained Facial Image Recognition and Retrieval

In this paper R-Theta Local Neighborhood Pattern (RTLNP) is proposed for facial image retrieval. RTLNP exploits relationships amongst the pixels in local neighborhood of the reference pixel at different angular and radial widths. The proposed encoding scheme divides the local neighborhood into sectors of equal angular width. These sectors are again divided into subsectors of two radial widths. Average grayscales values of these two subsectors are encoded to generate the micropatterns. Performance of the proposed descriptor has been evaluated and results are compared with the state of the art descriptors e.g. LBP, LTP, CSLBP, CSLTP, Sobel-LBP, LTCoP, LMeP, LDP, LTrP, MBLBP, BRINT and SLBP. The most challenging facial constrained and unconstrained databases, namely; AT&T, CARIA-Face-V5-Cropped, LFW, and Color FERET have been used for showing the efficiency of the proposed descriptor. Proposed descriptor is also tested on near infrared (NIR) face databases; CASIA NIR-VIS 2.0 and PolyU-NIRFD to explore its potential with respect to NIR facial images. Better retrieval rates of RTLNP as compared to the existing state of the art descriptors show the effectiveness of the descriptor

cs.CV↗

arXiv2022

Cascaded Asymmetric Local Pattern: A Novel Descriptor for Unconstrained Facial Image Recognition and Retrieval

Feature description is one of the most frequently studied areas in the expert systems and machine learning. Effective encoding of the images is an essential requirement for accurate matching. These encoding schemes play a significant role in recognition and retrieval systems. Facial recognition systems should be effective enough to accurately recognize individuals under intrinsic and extrinsic variations of the system. The templates or descriptors used in these systems encode spatial relationships of the pixels in the local neighbourhood of an image. Features encoded using these hand crafted descriptors should be robust against variations such as; illumination, background, poses, and expressions. In this paper a novel hand crafted cascaded asymmetric local pattern (CALP) is proposed for retrieval and recognition facial image. The proposed descriptor uniquely encodes relationship amongst the neighbouring pixels in horizontal and vertical directions. The proposed encoding scheme has optimum feature length and shows significant improvement in accuracy under environmental and physiological changes in a facial image. State of the art hand crafted descriptors namely; LBP, LDGP, CSLBP, SLBP and CSLTP are compared with the proposed descriptor on most challenging datasets namely; Caltech-face, LFW, and CASIA-face-v5. Result analysis shows that, the proposed descriptor outperforms state of the art under uncontrolled variations in expressions, background, pose and illumination.

cs.CV↗

arXiv2022

Centre Symmetric Quadruple Pattern: A Novel Descriptor for Facial Image Recognition and Retrieval

Facial features are defined as the local relationships that exist amongst the pixels of a facial image. Hand-crafted descriptors identify the relationships of the pixels in the local neighbourhood defined by the kernel. Kernel is a two dimensional matrix which is moved across the facial image. Distinctive information captured by the kernel with limited number of pixel achieves satisfactory recognition and retrieval accuracies on facial images taken under constrained environment (controlled variations in light, pose, expressions, and background). To achieve similar accuracies under unconstrained environment local neighbourhood has to be increased, in order to encode more pixels. Increasing local neighbourhood also increases the feature length of the descriptor. In this paper we propose a hand-crafted descriptor namely Centre Symmetric Quadruple Pattern (CSQP), which is structurally symmetric and encodes the facial asymmetry in quadruple space. The proposed descriptor efficiently encodes larger neighbourhood with optimal number of binary bits. It has been shown using average entropy, computed over feature images encoded with the proposed descriptor, that the CSQP captures more meaningful information as compared to state of the art descriptors. The retrieval and recognition accuracies of the proposed descriptor has been compared with state of the art hand-crafted descriptors (CSLBP, CSLTP, LDP, LBP, SLBP and LDGP) on bench mark databases namely; LFW, Colour-FERET, and CASIA-face-v5. Result analysis shows that the proposed descriptor performs well under controlled as well as uncontrolled variations in pose, illumination, background and expressions.

cs.MM↗

arXiv2021

Angle Based Feature Learning in GNN for 3D Object Detection using Point Cloud

In this paper, we present new feature encoding methods for Detection of 3D objects in point clouds. We used a graph neural network (GNN) for Detection of 3D objects namely cars, pedestrians, and cyclists. Feature encoding is one of the important steps in Detection of 3D objects. The dataset used is point cloud data which is irregular and unstructured and it needs to be encoded in such a way that ensures better feature encapsulation. Earlier works have used relative distance as one of the methods to encode the features. These methods are not resistant to rotation variance problems in Graph Neural Networks. We have included angular-based measures while performing feature encoding in graph neural networks. Along with that, we have performed a comparison between other methods like Absolute, Relative, Euclidean distances, and a combination of the Angle and Relative methods. The model is trained and evaluated on the subset of the KITTI object detection benchmark dataset under resource constraints. Our results demonstrate that a combination of angle measures and relative distance has performed better than other methods. In comparison to the baseline method(relative), it achieved better performance. We also performed time analysis of various feature encoding methods.

cs.CV↗