arXiv subjects

Sachin Grover

Publications and source records attributed to Sachin Grover.

15 recordsLinked to original sources

arXiv2026

Scattering of Null strings - Flipped Vacuum & CHY

We investigate the scattering of null tensionless strings. Classical null strings, given by the ILST action, give rise to three distinct quantum theories built on different vacua and representations of the underlying 2d Conformal Carroll algebra (CCA) which form the residual gauge symmetries on the null worldsheet. In this paper, we are interested in the quantum theory built out of the so-called ``flipped'' vacuum which realises the highest weight representation of the 2d CCA. By considering a novel class of vertex operators, intimately connected to the compactified null string, we build scattering amplitudes of the theory. We show that one naturally obtains the Cachazo-He-Yuan (CHY) formulae when one considers scattering of massless states in the null ``flipped'' string.

hep-th↗

arXiv2026

Double Copy from the Flipped Null String

We study the double-copy structure of tree amplitudes of the bosonic null string in the flipped vacuum. We show that its level-one vector correlator supplies the Cachazo-He-Yuan (CHY) kinematic half-integrand of the higher-derivative $(DF)^2$ gauge theory, while distinguished compact momentum-winding sectors generate the complementary Parke-Taylor factor. For factorized level-two states, the same Carrollian world-sheet produces two kinematic half-integrands and hence the symmetric double copy to the six-derivative Weyl-cubed graviton sector. Lattice sectors with common external momenta reproduce the Bern-Carrasco-Johansson (BCJ) ordering relation and the corresponding field-theory Kawai-Lewellen-Tye (KLT) representation. These results provide a direct Carrollian world-sheet origin of the double copy and point toward a current-algebra realization of the emergent color sector compatible with the $O(d,d)$ structure of the compact lattice.

hep-th↗

arXiv2026

High energy scattering and null strings

We propose an instrinsic worldsheet description of the ultra-high energy regime of string scattering based on worldsheet symmetries. At very high energies, the fundamental string becomes tensionless and in flat target spacetimes, the worldsheet becomes a null surface. Tensionless null strings thus emerge and the worldsheet symmetries morph from two copies of the Virasoro algebra to the two dimensional (2d) conformal Carroll or equivalently the 3d Bondi-van der Burgh-Metzner-Sachs (BMS) algebra. Tensionless strings have three inequivalent vacua over which they can be constructed, leading to distinct quantum theories. High energy tensile strings naturally connect to null strings built on the so-called induced vacuum. Our principle goal in this paper is the construction of scattering amplitudes for null strings in the induced vacuum. We show that these amplitudes, constructed from worldsheet methods of the null string, coincide with the high energy limit of usual string amplitudes. A crucial part of our analysis is the construction of integrated vertex operators. This achieved by relying on lessons from the parent string theory and following the tensionless limit carefully. A striking feature of the null string is the blurring of differences between open and closed strings. We see this at the level of the amplitudes as well. We then focus on four-point amplitudes and recover all expected regimes including the Gross-Mende regime and the Regge limit. We finally comment on a new class of vertex operators which arise naturally only in the zero tension string. This reproduces all our earlier analyses when put onshell but also has hints of signatures beyond the perturbative tensile string.

hep-th↗

arXiv2025

Prompted Policy Search: Reinforcement Learning through Linguistic and Numerical Reasoning in LLMs

Reinforcement Learning (RL) traditionally relies on scalar reward signals, limiting its ability to leverage the rich semantic knowledge often available in real-world tasks. In contrast, humans learn efficiently by combining numerical feedback with language, prior knowledge, and common sense. We introduce Prompted Policy Search (ProPS), a novel RL method that unifies numerical and linguistic reasoning within a single framework. Unlike prior work that augment existing RL components with language, ProPS places a large language model (LLM) at the center of the policy optimization loop-directly proposing policy updates based on both reward feedback and natural language input. We show that LLMs can perform numerical optimization in-context, and that incorporating semantic signals, such as goals, domain knowledge, and strategy hints can lead to more informed exploration and sample-efficient learning. ProPS is evaluated across fifteen Gymnasium tasks, spanning classic control, Atari games, and MuJoCo environments, and compared to seven widely-adopted RL algorithms (e.g., PPO, SAC, TRPO). It outperforms all baselines on eight out of fifteen tasks and demonstrates substantial gains when provided with domain knowledge. These results highlight the potential of unifying semantics and numerics for transparent, generalizable, and human-aligned RL.

cs.LG↗

arXiv2025

OpenThoughts: Data Recipes for Reasoning Models

Reasoning models have made rapid progress on many benchmarks involving math, code, and science. Yet, there are still many open questions about the best training recipes for reasoning since state-of-the-art models often rely on proprietary datasets with little to no public information available. To address this, the goal of the OpenThoughts project is to create open-source datasets for training reasoning models. After initial explorations, our OpenThoughts2-1M dataset led to OpenThinker2-32B, the first model trained on public reasoning data to match DeepSeek-R1-Distill-32B on standard reasoning benchmarks such as AIME and LiveCodeBench. We then improve our dataset further by systematically investigating each step of our data generation pipeline with 1,000+ controlled experiments, which led to OpenThoughts3. Scaling the pipeline to 1.2M examples and using QwQ-32B as teacher yields our OpenThoughts3-7B model, which achieves state-of-the-art results: 53% on AIME 2025, 51% on LiveCodeBench 06/24-01/25, and 54% on GPQA Diamond - improvements of 15.3, 17.2, and 20.5 percentage points compared to the DeepSeek-R1-Distill-Qwen-7B. All of our datasets and models are available on https://openthoughts.ai.

cs.LG↗

arXiv2024

Duality defects in $D_n$-type Niemeier lattice CFTs

We discuss the construction of duality defects in $c=24$ meromorphic CFTs that correspond to Niemeier lattices. We will illustrate our constructions for the $D_n$-type lattices. We will identify non-anomalous $\mathbb{Z}_2$ symmetries of these theories, and we show that on orbifolding with respect to these symmetries, these theories map to each other. We investigate this map, and in the case of self-dual orbifolds, we provide the duality defect partition functions. We show that exchange automorphisms in some CFTs give rise to a new class of defect partition functions.

hep-th↗

arXiv2023

Looking for the $G_2$ Higgs Branch of 4D Rank 1 SCFTs

The Schur index of the Higgs branch of 4-dimensional $\mathcal{N}=2$ SCFTs is related to the spectrum of non-unitary 2-dimensional CFTs. The rank 1 case has been shown to lead to the non-unitary CFTs with Deligne-Cvitanovic (DC) exceptional sequence of Lie groups. We show that a subsequence $(A_0, A_{\frac{1}{2}}, A_1, A_2, D_4)$ within the non-unitary sequence is related to a subsequence in the Mathur-Mukhi-Sen (MMS) sequence of unitary theories. We show that 2D non-unitary $G_2$ theory is related to unitary $E_6$ theory, and using this result along with the Galois conjugation, we propose that the $G_2$ Higgs branch is a sub-branch of the $E_6$ Higgs branch.

hep-th↗

arXiv2023

Quasi-Characters in $\widehat{su(2)}$ Current Algebra at Fractional Levels

We study the even characters of $\widehat{su(2)}$ conformal field theories (CFTs) at admissible fractional levels obtained from the difference of the highest weight characters in the unflavoured limit. We show that admissible even character vectors arise only in three special classes of admissible fractional levels which include the threshold levels, the positive half-odd integer levels, and the isolated level at -$5/4$. Among them, we show that the even characters of the half-odd integer levels map to the difference of characters of $\widehat{su(2)}_{4N+4}$, with $N\in\mathbb{Z}_{>0}$, although we prove that they do not correspond to rational CFTs. The isolated level characters maps to characters of two subsectors with $\widehat{so(5)}_1$ and $\widehat{su(2)}_1$ current algebras. Furthermore, for the $\widehat{su(2)}_1$ subsector of the isolated level, we introduce discrete flavour fugacities. The threshold levels saturate the admissibility bound and their even characters have previously been shown to be proportional to the unflavoured characters of integrable representations in $\widehat{su(2)}_{4N}$ CFTs, where $N\in\mathbb{Z}_{> 0}$ and we reaffirm this result. Except at the three classes of fractional levels, we find special inadmissible characters called quasi-characters which are nice vector valued modular functions but with $q$-series coefficients violating positivity but not integrality.

hep-th↗

arXiv2023

A Domain-Independent Agent Architecture for Adaptive Operation in Evolving Open Worlds

Model-based reasoning agents are ill-equipped to act in novel situations in which their model of the environment no longer sufficiently represents the world. We propose HYDRA - a framework for designing model-based agents operating in mixed discrete-continuous worlds, that can autonomously detect when the environment has evolved from its canonical setup, understand how it has evolved, and adapt the agents' models to perform effectively. HYDRA is based upon PDDL+, a rich modeling language for planning in mixed, discrete-continuous environments. It augments the planning module with visual reasoning, task selection, and action execution modules for closed-loop interaction with complex environments. HYDRA implements a novel meta-reasoning process that enables the agent to monitor its own behavior from a variety of aspects. The process employs a diverse set of computational methods to maintain expectations about the agent's own behavior in an environment. Divergences from those expectations are useful in detecting when the environment has evolved and identifying opportunities to adapt the underlying models. HYDRA builds upon ideas from diagnosis and repair and uses a heuristics-guided search over model changes such that they become competent in novel conditions. The HYDRA framework has been used to implement novelty-aware agents for three diverse domains - CartPole++ (a higher dimension variant of a classic control problem), Science Birds (an IJCAI competition problem), and PogoStick (a specific problem domain in Minecraft). We report empirical observations from these domains to demonstrate the efficacy of various components in the novelty meta-reasoning process.

cs.AI↗

arXiv2023

Heuristic Search For Physics-Based Problems: Angry Birds in PDDL+

This paper studies how a domain-independent planner and combinatorial search can be employed to play Angry Birds, a well established AI challenge problem. To model the game, we use PDDL+, a planning language for mixed discrete/continuous domains that supports durative processes and exogenous events. The paper describes the model and identifies key design decisions that reduce the problem complexity. In addition, we propose several domain-specific enhancements including heuristics and a search technique similar to preferred operators. Together, they alleviate the complexity of combinatorial search. We evaluate our approach by comparing its performance with dedicated domain-specific solvers on a range of Angry Birds levels. The results show that our performance is on par with these domain-specific approaches in most levels, even without using our domain-specific search enhancements.

cs.AI↗

arXiv2021

COVID-19 India Dataset: Parsing COVID-19 Data in Daily Health Bulletins from States in India

While India has been one of the hotspots of COVID-19, data about the pandemic from the country has proved to be largely inaccessible at scale. Much of the data exists in unstructured form on the web, and limited aspects of such data are available through public APIs maintained manually through volunteer effort. This has proved to be difficult both in terms of ease of access to detailed data and with regards to the maintenance of manual data-keeping over time. This paper reports on our effort at automating the extraction of such data from public health bulletins with the help of a combination of classical PDF parsers and state-of-the-art machine learning techniques. In this paper, we will describe the automated data-extraction technique, the nature of the generated data, and exciting avenues of ongoing work.

cs.CL↗

arXiv2021

Characterizing Human Explanation Strategies to Inform the Design of Explainable AI for Building Damage Assessment

Explainable AI (XAI) is a promising means of supporting human-AI collaborations for high-stakes visual detection tasks, such as damage detection tasks from satellite imageries, as fully-automated approaches are unlikely to be perfectly safe and reliable. However, most existing XAI techniques are not informed by the understandings of task-specific needs of humans for explanations. Thus, we took a first step toward understanding what forms of XAI humans require in damage detection tasks. We conducted an online crowdsourced study to understand how people explain their own assessments, when evaluating the severity of building damage based on satellite imagery. Through the study with 60 crowdworkers, we surfaced six major strategies that humans utilize to explain their visual damage assessments. We present implications of our findings for the design of XAI methods for such visual detection contexts, and discuss opportunities for future research.

cs.HC↗

arXiv2020

Model Elicitation through Direct Questioning

The future will be replete with scenarios where humans are robots will be working together in complex environments. Teammates interact, and the robot's interaction has to be about getting useful information about the human's (teammate's) model. There are many challenges before a robot can interact, such as incorporating the structural differences in the human's model, ensuring simpler responses, etc. In this paper, we investigate how a robot can interact to localize the human model from a set of models. We show how to generate questions to refine the robot's understanding of the teammate's model. We evaluate the method in various planning domains. The evaluation shows that these questions can be generated offline, and can help refine the model through simple answers.

cs.AI↗

arXiv2018

Plan Explanations as Model Reconciliation -- An Empirical Study

Recent work in explanation generation for decision making agents has looked at how unexplained behavior of autonomous systems can be understood in terms of differences in the model of the system and the human's understanding of the same, and how the explanation process as a result of this mismatch can be then seen as a process of reconciliation of these models. Existing algorithms in such settings, while having been built on contrastive, selective and social properties of explanations as studied extensively in the psychology literature, have not, to the best of our knowledge, been evaluated in settings with actual humans in the loop. As such, the applicability of such explanations to human-AI and human-robot interactions remains suspect. In this paper, we set out to evaluate these explanation generation algorithms in a series of studies in a mock search and rescue scenario with an internal semi-autonomous robot and an external human commander. We demonstrate to what extent the properties of these algorithms hold as they are evaluated by humans, and how the dynamics of trust between the human and the robot evolve during the process of these interactions.

cs.AI↗

arXiv2017

Texture Synthesis with Recurrent Variational Auto-Encoder

We propose a recurrent variational auto-encoder for texture synthesis. A novel loss function, FLTBNK, is used for training the texture synthesizer. It is rotational and partially color invariant loss function. Unlike L2 loss, FLTBNK explicitly models the correlation of color intensity between pixels. Our texture synthesizer generates neighboring tiles to expand a sample texture and is evaluated using various texture patterns from Describable Textures Dataset (DTD). We perform both quantitative and qualitative experiments with various loss functions to evaluate the performance of our proposed loss function (FLTBNK) --- a mini-human subject study is used for the qualitative evaluation.

cs.CV↗