arXiv subjects

Debarshi Banerjee

Publications and source records attributed to Debarshi Banerjee.

13 recordsLinked to original sources

arXiv2026

REBASE: Reference-Background Subspace Elimination for Training-Free In-Context Segmentation

Training-free in-context segmentation enables new object categories to be introduced at inference time from a single annotated reference image, eliminating the retraining and memory overhead of class-incremental learning. Recent approaches achieve this by combining vision foundation models for semantic correspondence with promptable segmentation networks like SAM. However, their performance is fundamentally limited by the quality of the cross-image similarity map; shared contextual backgrounds between the reference and query systematically elevate similarity in non-target regions, degrading prompt localization. We present REBASE, a training-free framework that explicitly suppresses these spurious contextual correspondences. Our method identifies the low-rank background feature subspace from the reference image and project the reference and query features onto its orthogonal complement in closed form, yielding cleaner semantic matching. We then generate positive point prompts using similarity-weighted farthest-point sampling, paired with a refined dense similarity prior. Without any training or parameter updates, our approach establishes a new state of the art among training-free methods on PACO-Part, FSS-1000, and cross-domain datasets such as ISIC2018, demonstrating that explicit background subspace removal is a highly effective principle for one-shot localization.

cs.CV↗

arXiv2026

Investigating causality between principal components in protein dynamics

Principal component analysis (PCA) is widely used to characterize collective protein motions from molecular dynamics (MD) simulations. While PCA identifies the dominant modes of structural fluctuation, it does not reveal whether different principal components (PCs) causally influence each other. Here, we investigate this question using a recently introduced causal-discovery framework [Del Tatto et al, PNAS 2024], which allows to infer putative causal asymmetries between high-dimensional time series. We apply this approach to long-timescale MD trajectories of two proteins. By analyzing relationships among PCs, we construct directed networks describing how PCs influence one another across time scales. These directional relationships, whose existence is a necessary condition for the presence of a causal link, are not captured by conventional covariance-based analyses and provide information that is complementary to PCA and Time-lagged Independent Component Analysis (TICA). Our results suggest that our causal inference approach can uncover previously hidden aspects of the dynamical organization of protein motions and offer a new perspective on this very popular class of collective variables.

physics.chem-ph↗

arXiv2026

Machine learning the non-radiative decay modes in photochemical processes

Non-radiative decay in photoexcited molecular systems is driven by nuclear motion toward conical intersections (CIs), where electronic states become degenerate and nonadiabatic transitions occur. Identifying the nuclear degrees of freedom responsible for CI access from nonadiabatic molecular dynamics (NAD) simulations remains challenging because the underlying motions are high-dimensional and collective. Here, we introduce an unsupervised, information-theoretic framework based on Differentiable Information Imbalance (DII) to identify the nuclear coordinates governing CI access directly from trajectory surface hopping (TSH) simulations. By quantifying correlations between structural descriptors and electronic observables, including energy gaps and oscillator strengths, the method ranks nuclear degrees of freedom by predictive relevance. A multi-step protocol then extracts low-dimensional, physically interpretable modes associated with non-radiative decay. We apply the framework to the methaniminium cation, furan, L-glutamine, L-pyroglutamine-ammonium, and a photoactive molecular motor. Across all systems, the method recovers known mechanistic coordinates while revealing the relative importance of competing modes when multiple structural distortions contribute to CI access. The analysis also reveals a systematic distinction between observables: energy gaps are typically governed by a small number of localized coordinates, whereas oscillator strengths depend on more collective and distributed structural rearrangements. Overall, the DII-based framework combines predictive power with direct interpretability, providing a general and scalable route for extracting mechanistic insight from high-dimensional NAD data and constructing reduced-dimensional models of excited-state dynamics.

physics.chem-ph↗

arXiv2026

Causality in Liquid Water as a Hallmark of Emergent Glassy Dynamics

In molecular liquids such as water, time-delayed influences between microscopic or mesoscopic variables are typically probed using time-correlation functions, which are symmetric under detailed balance and therefore blind to dynamical asymmetries. Here, we characterize waters dynamics using a causal inference metric that captures asymmetric couplings between collective variables. Analyzing equilibrium molecular dynamics simulations at ambient conditions and in the high-density liquid (HDL) regime of supercooled water, we uncover pronounced asymmetries in the couplings between orientational and translational degrees of freedom across multiple time and length scales. At room temperature, rotational modes remain largely decoupled from translations. In contrast, in the supercooled HDL regime, translational motions emerge as the primary drivers of the dynamics, suggesting facilitation-like relaxation mechanisms characteristic of glassy systems. These results reveal a qualitative reorganization of dynamical couplings across thermodynamic conditions, implying that molecular liquids at thermal equilibrium can exhibit an emergent directionality in their fluctuation couplings. As a consequence, our analysis reveals that external perturbations acting on specific degrees of freedom can induce a stronger arrow of time in the causal relations between translational and orientational modes.

physics.chem-ph↗

arXiv2026

GenSeg-R1: RL-Driven Vision-Language Grounding for Fine-Grained Referring Segmentation

We study fine-grained referring image segmentation via a decoupled reason-then-segment pipeline. A vision-language model (VLM) receives an image and a natural-language query, reasons about the scene, and emits structured spatial prompts: a bounding box plus two interior keypoints for every referred instance. A frozen promptable segmenter (SAM 2) converts these prompts into high-quality masks. Within our GenSeg-R1 framework we finetune Qwen3-VL models (4B and 8B parameters) using Group Relative Policy Optimization (GRPO), requiring no supervised reasoning-chain annotations. On RefCOCOg validation our best model (GenSeg-R1-8B) achieves 0.7127 cIoU and 0.7382 mIoU, substantially outperforming the corresponding Qwen3-VL Instruct baselines (+15.3 and +21.9 points, respectively) and surpassing Seg-Zero-7B [3] by +3.3 cIoU under identical evaluation. We further introduce GenSeg-R1-G, a variant trained on GRefCOCO [9] with a SAM 2 in-the-loop reward that directly optimizes mask quality. On GRefCOCO validation GenSeg-R1-G achieves 76.69% target mIoU with 82.40% accuracy on negative (no-target) prompts, substantially outperforming Seg-R1-7B and Seg-Zero-7B, which lack no-target detection capability. On ReasonSeg test, GenSeg-R1-4B reaches 68.40% mIoU, surpassing Seg-Zero-7B by +7.0 and Seg-R1-7B by +10.7 points.

cs.CV↗

arXiv2025

State-of-the-art Small Language Coder Model: Mify-Coder

We present Mify-Coder, a 2.5B-parameter code model trained on 4.2T tokens using a compute-optimal strategy built on the Mify-2.5B foundation model. Mify-Coder achieves comparable accuracy and safety while significantly outperforming much larger baseline models on standard coding and function-calling benchmarks, demonstrating that compact models can match frontier-grade models in code generation and agent-driven workflows. Our training pipeline combines high-quality curated sources with synthetic data generated through agentically designed prompts, refined iteratively using enterprise-grade evaluation datasets. LLM-based quality filtering further enhances data density, enabling frugal yet effective training. Through disciplined exploration of CPT-SFT objectives, data mixtures, and sampling dynamics, we deliver frontier-grade code intelligence within a single continuous training trajectory. Empirical evidence shows that principled data and compute discipline allow smaller models to achieve competitive accuracy, efficiency, and safety compliance. Quantized variants of Mify-Coder enable deployment on standard desktop environments without requiring specialized hardware.

cs.SE↗

arXiv2025

Investigating the Role of pH and Counterions in the Intrinsic Fluorescence of Solid-State L-Lysine

There is currently a growing interest in understanding the origins of intrinsic fluorescence as a way to design non-invasive probes for biophysical processes. In this regard, understanding how pH influences fluorescence in non-aromatic biomolecular assemblies is key to controlling their optical properties in realistic cellular conditions. Here, we combine experiments and theory to investigate the pH-dependent emission of solid-state L-Lysine (Lys). Lys aggregates prepared at different pH values using HCl and H$_2$SO$_4$ exhibit protonation- and counterion-dependent morphology and fluorescence, as shown by microscopy and steady-state measurements. We find an enhancement in the fluorescence moving from acidic to basic conditions. To uncover the molecular origin of these trends, we performed non-adiabatic molecular dynamics simulations on three Lys crystal models representing distinct protonation states. Our simulations indicate that enhanced protonation under acidic conditions facilitates non-radiative decay via proton transfer, whereas basic conditions favor radiative decay. Our combined experimental-theoretical work highlights pH and counterion identity as key factors tuning fluorescence in Lys assemblies, offering insights for designing pH responsive optical materials based on non-aromatic amino acids.

physics.chem-ph↗

arXiv2025

Deconstructing the Origins of Interfacial Catalysis: Why Electric Fields are Inseparable from Solvation

In the last decade, there has been a surge of experiments showing that certain chemical reactions undergo an enormous boost when taken from bulk aqueous conditions to microdroplet environments. The microscopic basis of this phenomenon remains elusive and continues to be widely debated. One of the key driving forces invoked are the specific properties of the air-water interface including the presence of large electric fields and distinct solvation at the surface. Here, using a combination of classical molecular dynamics simulations, the chemical physics of solvation, and unsupervised learning approaches, we place these assumptions under close scrutiny. Using phenol as a model system, we demonstrate that the electric field at the surface of water is not anomalous or unique compared to bulk water conditions. Furthermore, the electric field fluctuations de-correlate on a timescale of ~10 ps implying that their role in activating much slower chemical reactions remains inconclusive. We deploy a recently developed unsupervised learning approach, dubbed information balance, which detects in an agnostic fashion the relationship between the electric field and solvation collective variables. It turns out that the electric field on the hydroxyl group of the phenol is mostly determined by phenol hydration including the proximity and orientation of nearby water molecules. We caution that the growing attention of the role that electric fields have garnered in enhancing chemical reactivity at the air-water interface, may not reflect their actual importance.

physics.chem-ph↗

arXiv2025

Towards a robust approach to infer causality in molecular systems satisfying detailed balance

The ability to distinguish between correlation and causation of variables in molecular systems remains an interesting and open area of investigation. In this work, we probe causality in a molecular system using two independent computational methods that infer the causal direction through the language of information transfer. Specifically, we demonstrate that a molecular dynamics simulation involving a single Tryptophan in liquid water displays asymmetric information transfer between specific collective variables, such as solute and solvent coordinates. Analyzing a discrete Markov-state and Langevin dynamics on a 2D free energy surface, we show that the same kind of asymmetries can emerge even in extremely simple systems, undergoing equilibrium and time-reversible dynamics. We use these model systems to rationalize the unidirectional information transfer in the molecular system in terms of asymmetries in the underlying free energy landscape and/or relaxation dynamics of the relevant coordinates. Finally, we propose a computational experiment that allows one to decide if an asymmetric information transfer between two variables corresponds to a genuine causal link.

physics.chem-ph↗

arXiv2025

Turning on the Light: Polymorphism-Induced Photoluminescence in Cysteine Crystals

Photoluminescence of non-aromatic supramolecular chemical assemblies has attracted considerable attention in recent years due to its potential for use in molecular sensing and imaging technologies. The underlying structural origins, the mechanisms of light emission in these systems, and the generality of this phenomenon remain elusive. Here, we demonstrate that crystals of L-Cysteine (Cys) formed in heavy water ($\text{D}_2\text{O}$) exhibit distinct packing and hydrogen-bond networks, resulting in significantly enhanced photoluminescence compared to those prepared in $\text{H}_2\text{O}$. Using advanced excited-state simulations, we elucidate the nature of electronic transitions that activate vibrational modes of Cys in $\text{H}_2\text{O}$, particularly those involving thiol (S-H) and amine (C-N) groups, which lead to non-radiative decay. For the crystal formed in $\text{D}_2\text{O}$, these modes appear to be more constrained, and we also observe intersystem crossing from the singlet to the triplet state, indicating a potentially more complex light emission mechanism. Our findings provide new insights into this intriguing phenomenon and introduce innovative design principles for generating emergent fluorophores.

physics.chem-ph↗

arXiv2024

Trajectory Surface Hopping with Tight Binding Density Functional Theory applied to Molecular Motors

Non-adiabatic molecular dynamics (NAMD) has become an essential computational technique for studying the photophysical relaxation of molecular systems after light absorption. These phenomena require approximations that go beyond the Born-Oppenheimer approximation, and the accuracy of the results heavily depends on the electronic structure theory employed. Sophisticated electronic methods, however, make these techniques computationally expensive, even for medium size systems. Consequently, simulations are often performed on simplified models to interpret experimental results. In this context, a variety of techniques have been developed to perform NAMD using approximate methods, particularly Density Functional Tight Binding (DFTB). Despite the use of these techniques on large systems where ab initio methods are computationally prohibitive, a comprehensive validation has been lacking. In this work, we present a new implementation of trajectory surface hopping (TSH) combined with DFTB, utilizing non-adiabatic coupling vectors (NACVs). We selected two different systems for validation, providing an exhaustive comparison with higher-level electronic structure methods. As a case study, we simulated a system from the class of molecular motors, which has been extensively studied experimentally but remains challenging to simulate with ab initio methods due to its inherent complexity. Our approach effectively captures the key photophysical mechanism of dihedral rotation after absorption of light. Additionally, we successfully reproduce the transition from the bright to dark states observed in the time dependent fluorescence experiments, providing valuable insights into this critical part of the photophysical behavior in molecular motors.

physics.chem-ph↗

arXiv2024

Aqueous Solution Chemistry In Silico and the Role of Data Driven Approaches

The use of computer simulations to study the properties of aqueous systems is, today more than ever, an active area of research. In this context, during the last decade there has been a tremendous growth in the use of data-driven approaches to develop more accurate potentials for water as well as to characterize its complexity in chemical and biological contexts. We highlight the progress, giving a historical context, on the path to the development of many-body and reactive potentials to model aqueous chemistry, including the role of machine learning strategies. We focus specifically on conceptual and methodological challenges along the way in performing simulations that seek to tackle problems in modeling the chemistry of aqueous solutions. In conclusion, we summarize our perspectives on the use and integration of advanced data-science techniques to provide chemical insights in physical chemistry and how this will influence computer simulations of aqueous systems in the future.

physics.chem-ph↗

arXiv2024

Exploring the Mechanisms Behind Non Aromatic Fluorescence with Density Functional Tight Binding Method

Recent experimental findings reveal non-conventional fluorescence emission in biological systems devoid of conjugated bonds or aromatic compounds, termed \textit{Non-Aromatic Fluorescence} (NAF). This phenomenon is exclusive to aggregated or solid states, remaining absent in monomeric solutions. Previous studies focused on small model systems in vacuum show that the carbonyl stretching mode along with strong interaction of short hydrogen bonds (SHBs) remain the primary vibrational mode explaining NAF in these systems. In order to simulate larger model systems taking into account the effects of the surrounding environment, in this work we propose using the density functional tight-binding (DFTB) method in combination with non-adiabatic molecular dynamics (NAMD) and the mixed quantum/molecular mechanics (QM/MM) approach. We investigate the mechanism behind NAF in the crystal structure of L-pyroglutamine-ammonium, comparing it with the related non-fluorescent amino acid L-glutamine. Our results extend our previous findings to more realistic systems, demonstrating the efficiency and robustness of the proposed DFTB method in the context of NAMD in biological systems. Furthemore, due to its inherent low computational cost, this method allows for a better sampling on the non-radiative events at the conical intersection which is crucial for a complete understanding of this phenomenon. Beyond contributing to the ongoing exploration of NAF, this work paves the way for future application of this method in more complex biological systems such as amyloid aggregates, biomaterials and non-aromatic proteins.

physics.chem-ph↗