arXiv subjects

Harsh Raj

Publications and source records attributed to Harsh Raj.

At least 19 recordsLinked to original sources

arXiv2026

Harbor Adapters and Harbor-Index: Infrastructure and a Curated Meta-Dataset for Large-Scale Agentic Evaluation

Evaluating agents on the growing number of agentic benchmarks is challenging because they often require complex environments and agent integrations. We introduce Harbor Adapters, a unified evaluation infrastructure for agentic benchmarks. Our work makes three contributions. First, we develop benchmark adapters that port more than 80 benchmarks to evaluate arbitrary agents, and validate them through rigorous code review and parity experiments. Second, we conduct a large-scale evaluation of 8 models spanning capability tiers across 54 benchmarks; every model is run with Terminus-2 and with one of 3 native harnesses. This enables a broader analysis of agent capabilities and failure modes than was previously possible. Third, we introduce Harbor-Index, a curated set of 82 difficult, diverse, and high-quality tasks spanning 29 benchmarks, refined from the adapted suite through difficulty filtering, AI and human audit, and an audit-and-fix loop. Harbor-Index preserves the challenge and breadth of large-scale agentic evaluations while being affordable to run; no evaluated model-harness configuration exceeds 30% pass rate, and the strongest (GPT-5.5 with Codex) reaches 28.0%. We release the adapters, evaluation results, in-depth analysis, and Harbor-Index as open-source artifacts to support more reliable and comprehensive evaluation of language-model agents.

cs.AI↗

arXiv2026

Synthetic Persona Pretraining: Alignment from Token Zero

As language-model-based AI is increasingly deployed in autonomous settings, aligning its goals and values with those of humans becomes critical. Today, alignment, and the assistant identity itself, are typically introduced only after pretraining, once behavioral priors are already established. This can make values a thin overlay, rather than deeply rooted, and facilitate subsequent misalignment. Pursuing a different paradigm, we introduce Synthetic Persona Pretraining (SPP), which installs the desired assistant persona from token zero in pretraining. First, we annotate pretraining documents with value-aligned first-person reflections derived from a normative value constitution. Second, we pretrain via the standard cross-entropy loss on standard pretraining documents as well as their reflections, which installs the desired persona among a multitude of other personas. Finally, we post-train on user-assistant dialogue data, which binds this desired persona to the assistant identity, a process we call persona binding. By pretraining models up to 3B parameters on 500B tokens, we show that SPP improves constitution following and jailbreak robustness, and reduces the misalignment rate in out-of-distribution moral dilemmas, while preserving capabilities. Early intervention matters: compared with alignment from token zero, introducing SPP only at the end of pretraining yields weaker constitution adherence, does not shift value priorities, and leads to less aligned choices in dilemmas. This advantage depends on persona binding and, importantly, increases with pretraining budget. Overall, our results show that shaping values early is critical for alignment and establish pretraining-time persona interventions as an effective approach to do so.

cs.LG↗

arXiv2026

Model or Harness? An Interaction-Centric Taxonomy for Localizing Agent Failures

Existing evaluations often reduce agent failures to system-level outcomes, obscuring where the fault originated and which intervention would improve the agent system. This creates a repair-assignment problem: the same visible failure may call for model post-training, harness engineering, environment redesign, or benchmark repair depending on its source. Because agent behavior emerges from interactions among models, harnesses, users, tools, memory, and environments, outcome-level labels are often insufficient for improvement. Most failure taxonomies do little to resolve this problem because they are benchmark-specific and lack a shared structure. We introduce an interaction-centric taxonomy that localizes failures to the interactions in which they originate and identifies the responsible component. It organizes 41 failure modes by assigning each to an edge between two components and a fault side indicating where the repair belongs. This makes the taxonomy actionable: model-side failures identify targets for post-training, harness-side failures point to scaffolding and tool-integration fixes, and environment or grader failures reveal evaluation conditions requiring redesign. The schema applies across agent architectures, from coding assistants to long-horizon personal assistants and multi-agent systems. We ground the taxonomy in worked examples from public benchmarks, model system cards, published reports, and logged agent trajectories, and evaluate its reproducibility using independent reasoning agents as judges. Across four frontier models, the strongest judge reaches Cohen's $\kappa=0.76$ against human category labels, suggesting that the categories capture shared structure rather than annotator-specific preferences.

cs.AI↗

arXiv2026

OpenThoughts-Agent: Data Recipes for Agentic Models

Agentic language models dramatically expand the applications of AI yet little is publicly known about how to curate training data for broadly capable agents. Existing open efforts such as SWE-Smith, SERA, and Nemotron-Terminal typically target a single benchmark, leaving open the question of how to train models that generalize across diverse agentic tasks. The OpenThoughts-Agent (OT-Agent) project addresses this gap with a fully open data curation pipeline for training agentic models. We conduct more than 100 controlled ablation experiments to systematically investigate each stage of the pipeline, yielding insights on the importance of task sources and diversity. We then assemble a training set of 100K examples from our pipeline and fine-tune Qwen3-32B on this dataset, which yields an average accuracy of 44.8% across seven agentic benchmarks and a 3.9 percentage point improvement over the strongest existing open data agentic model (Nemotron-Terminal-32B, 40.9%). Moreover, our training data exhibits strong scaling properties, outperforming alternative open datasets at every training set size in compute-controlled comparisons. We publicly release our training sets, data pipeline, experimental data, and models at openthoughts.ai to support future open research on agentic model training.

cs.AI↗

arXiv2026

Joint Antenna Placement and Power Allocation for RSMA-Enabled Pinching Antenna Systems

This letter investigates a rate-splitting multiple access (RSMA)-enabled multi-user pinching antenna system (PASS). A fairness-aware sum-rate maximization problem is formulated to jointly optimize pinching antenna locations and common/private stream power allocation. The resulting mixed discrete-continuous non-convex problem is addressed using an alternating optimization framework that combines greedy antenna placement with successive convex approximation (SCA)-based power allocation. Numerical results demonstrate that the proposed RSMA-enabled PASS significantly improves achievable sum-rate, user fairness, and bit error rate (BER) performance compared with conventional non-RSMA PASS schemes.

eess.SP↗

arXiv2026

Consistency as a Testable Property: Statistical Methods to Evaluate AI Agent Reliability

This paper establishes a rigorous measurement science for AI agent reliability, providing a foundational framework for quantifying consistency under semantically preserving perturbations. By leveraging $U$-statistics for output-level reliability and kernel-based metrics for trajectory-level stability, we offer a principled approach to evaluating agents across diverse operating conditions. Our proposal highlights the important distinction between the core capability and execution robustness of an agent, showing that minor task-level variations can induce complete strategy breakdowns despite the agent possessing the requisite knowledge for the task. We validate our framework through extensive experiments on three agentic benchmarks, demonstrating that trajectory-level consistency metrics provide far greater diagnostic sensitivity than traditional pass@1 rates. By providing the mathematical tools to isolate where and why agents deviate, we enable the identification and rectification of architectural concerns that hinder the deployment of agents in high-stakes, real-world environments.

cs.AI↗

arXiv2026

Terminal-Bench: Benchmarking Agents on Hard, Realistic Tasks in Command Line Interfaces

AI agents may soon become capable of autonomously completing valuable, long-horizon tasks in diverse domains. Current benchmarks either do not measure real-world tasks, or are not sufficiently difficult to meaningfully measure frontier models. To this end, we present Terminal-Bench 2.0: a carefully curated hard benchmark composed of 89 tasks in computer terminal environments inspired by problems from real workflows. Each task features a unique environment, human-written solution, and comprehensive tests for verification. We show that frontier models and agents score less than 65\% on the benchmark and conduct an error analysis to identify areas for model and agent improvement. We publish the dataset and evaluation harness to assist developers and researchers in future work at https://www.tbench.ai/ .

cs.SE↗

arXiv2025

Analysis of Traffic Congestion in North Campus, Delhi University Using Continuous Time Models

This project investigates traffic congestion within North Campus, Delhi University (DU), using continuous time simulations implemented in UXSim to model vehicle movement and interaction. The study focuses on several key intersections, identifies recurring congestion points, and evaluates the effectiveness of conventional traffic management measures. Implementing signal timing optimization and modest intersection reconfiguration resulted in measurable improvements in simulated traffic flow. The results provide practical insights for local traffic management and illustrate the value of continuous time simulation methods for informing short-term interventions and longer-term planning.

eess.SY↗

arXiv2025

MixtureVitae: Open Web-Scale Pretraining Dataset With High Quality Instruction and Reasoning Data Built from Permissive-First Text Sources

We present MixtureVitae, an open-access pretraining corpus built to minimize legal risk while providing strong downstream performance. MixtureVitae follows a permissive-first, risk-mitigated sourcing strategy that combines public-domain and permissively licensed text (e.g., CC-BY/Apache) with carefully justified low-risk additions (e.g., government works and EU TDM-eligible sources). MixtureVitae adopts a simple, single-stage pretraining recipe that integrates a large proportion of permissive synthetic instruction and reasoning data-signals typically introduced during post-training and generally scarce in permissive web corpora. We categorize all sources into a three-tier scheme that reflects varying risk levels and provide shard-level provenance metadata to enable risk-aware usage. In controlled experiments using the open-sci-ref training protocol (fixed architectures and hyperparameters; 50B and 300B token budgets across 130M-1.7B parameters), models trained on MixtureVitae consistently outperform other permissive datasets across a suite of standard benchmarks, and at the 1.7B-parameters/300B-tokens setting, they surpass FineWeb-Edu and approach DCLM late in training. Performance is particularly strong on MMLU and on math and code benchmarks: a 1.7B model pretrained on 300B MixtureVitae tokens matches or exceeds a strong 1.7B instruction-tuned baseline on GSM8K, HumanEval, and MBPP, despite using over 36 times fewer tokens (300B vs. ~11T). Supported by a thorough decontamination analysis, these results show that permissive-first data with high instruction and reasoning density, tiered by licensing and provenance-related risk, can provide a practical and risk-mitigated foundation for training capable LLMs, reducing reliance on broad web scrapes without sacrificing competitiveness. Code: https://github.com/ontocord/mixturevitae

cs.CL↗

arXiv2025

Improving Consistency in Large Language Models through Chain of Guidance

Consistency is a fundamental dimension of trustworthiness in Large Language Models (LLMs). For humans to be able to trust LLM-based applications, their outputs should be consistent when prompted with inputs that carry the same meaning or intent. Despite this need, there is no known mechanism to control and guide LLMs to be more consistent at inference time. In this paper, we introduce a novel alignment strategy to maximize semantic consistency in LLM outputs. Our proposal is based on Chain of Guidance (CoG), a multistep prompting technique that generates highly consistent outputs from LLMs. For closed-book question-answering (Q&A) tasks, when compared to direct prompting, the outputs generated using CoG show improved consistency. While other approaches like template-based responses and majority voting may offer alternative paths to consistency, our work focuses on exploring the potential of guided prompting. We use synthetic data sets comprised of consistent input-output pairs to fine-tune LLMs to produce consistent and correct outputs. Our fine-tuned models are more than twice as consistent compared to base models and show strong generalization capabilities by producing consistent outputs over datasets not used in the fine-tuning process.

cs.CL↗

arXiv2024

Evaluating Defences against Unsafe Feedback in RLHF

While there has been progress towards aligning Large Language Models (LLMs) with human values and ensuring safe behaviour at inference time, safety guards can easily be removed when fine tuned on unsafe and harmful datasets. While this setting has been treated extensively, another popular training paradigm, learning from unsafe feedback with reinforcement learning, has previously been unexplored. This is concerning due to the widespread deployment of feedback collection systems. We address this gap by providing an analysis of learning settings where feedback is harmful, i.e. that unsafe samples are preferred over safe ones despite model developers goal to maintain safety. We find that safety-aligned LLMs easily explore unsafe action spaces via generating harmful text and optimize for reward that violates safety constraints indicating that current safety guards are not enough to prevent learning from unsafe feedback. In order to protect against this vulnerability, we adapt a number of both "implict" and "explicit" harmful fine-tuning defences to evaluate whether they are effective as learning constraints in an RLHF setting finding that no method is generally effective pointing to the need for more defence research. We end the paper with the observation that some defences work by performing "harmless reward hacking" for which we provide a theoretical explanation drawn from the theory of Constrained Markov Decision Processes and provide some direction for future defence development.

cs.LG↗

arXiv2024

On the Performance of IRS-Assisted SSK and RPM over Rician Fading Channels

This paper presents the index modulation, that is, the space-shift keying (SSK) and reflection phase modulation (RPM) schemes for intelligent reflecting surface (IRS)-assisted wireless network. IRS simultaneously reflects the incoming information signal from the base station and explicitly encodes the local information bits in the reflection phase shift of IRS elements. The phase shift of the IRS elements is employed according to local data from the RPM constellation. A joint detection using a maximum-likelihood (ML) decoder is performed for the SSK and RPM symbols over a realistic fading scenario modeled as the Rician fading channel. The pairwise error probability over Rician fading channels is derived and utilized to determine the average bit error rate. In addition, the ergodic capacity of the presented system is derived. The derived analytical results are verified and are in exact agreement with Monte-Carlo simulations.

cs.IT↗

arXiv2023

Semantic Consistency for Assuring Reliability of Large Language Models

Large Language Models (LLMs) exhibit remarkable fluency and competence across various natural language tasks. However, recent research has highlighted their sensitivity to variations in input prompts. To deploy LLMs in a safe and reliable manner, it is crucial for their outputs to be consistent when prompted with expressions that carry the same meaning or intent. While some existing work has explored how state-of-the-art LLMs address this issue, their evaluations have been confined to assessing lexical equality of single- or multi-word answers, overlooking the consistency of generative text sequences. For a more comprehensive understanding of the consistency of LLMs in open-ended text generation scenarios, we introduce a general measure of semantic consistency, and formulate multiple versions of this metric to evaluate the performance of various LLMs. Our proposal demonstrates significantly higher consistency and stronger correlation with human evaluations of output consistency than traditional metrics based on lexical consistency. Finally, we propose a novel prompting strategy, called Ask-to-Choose (A2C), to enhance semantic consistency. When evaluated for closed-book question answering based on answer variations from the TruthfulQA benchmark, A2C increases accuracy metrics for pretrained and finetuned LLMs by up to 47%, and semantic consistency metrics for instruction-tuned models by up to 7-fold.

cs.CL↗

arXiv2022

Measuring Reliability of Large Language Models through Semantic Consistency

While large pretrained language models (PLMs) demonstrate incredible fluency and performance on many natural language tasks, recent work has shown that well-performing PLMs are very sensitive to what prompts are feed into them. Even when prompts are semantically identical, language models may give very different answers. When considering safe and trustworthy deployments of PLMs we would like their outputs to be consistent under prompts that mean the same thing or convey the same intent. While some work has looked into how state-of-the-art PLMs address this need, they have been limited to only evaluating lexical equality of single- or multi-word answers and do not address consistency of generative text sequences. In order to understand consistency of PLMs under text generation settings, we develop a measure of semantic consistency that allows the comparison of open-ended text outputs. We implement several versions of this consistency metric to evaluate the performance of a number of PLMs on paraphrased versions of questions in the TruthfulQA dataset, we find that our proposed metrics are considerably more consistent than traditional metrics embodying lexical consistency, and also correlate with human evaluation of output consistency to a higher degree.

cs.CL↗

arXiv2022

AskYourDB: An end-to-end system for querying and visualizing relational databases using natural language

Querying databases for the right information is a time consuming and error-prone task and often requires experienced professionals for the job. Furthermore, the user needs to have some prior knowledge about the database. There have been various efforts to develop an intelligence which can help business users to query databases directly. However, there has been some successes, but very little in terms of testing and deploying those for real world users. In this paper, we propose a semantic parsing approach to address the challenge of converting complex natural language into SQL and institute a product out of it. For this purpose, we modified state-of-the-art models, by various pre and post processing steps which make the significant part when a model is deployed in production. To make the product serviceable to businesses we added an automatic visualization framework over the queried results.

cs.DB↗

arXiv2022

On Transfer of Adversarial Robustness from Pretraining to Downstream Tasks

As large-scale training regimes have gained popularity, the use of pretrained models for downstream tasks has become common practice in machine learning. While pretraining has been shown to enhance the performance of models in practice, the transfer of robustness properties from pretraining to downstream tasks remains poorly understood. In this study, we demonstrate that the robustness of a linear predictor on downstream tasks can be constrained by the robustness of its underlying representation, regardless of the protocol used for pretraining. We prove (i) a bound on the loss that holds independent of any downstream task, as well as (ii) a criterion for robust classification in particular. We validate our theoretical results in practical applications, show how our results can be used for calibrating expectations of downstream robustness, and when our results are useful for optimal transfer learning. Taken together, our results offer an initial step towards characterizing the requirements of the representation function for reliable post-adaptation performance.

cs.LG↗

arXiv2022

GANDALF: Gated Adaptive Network for Deep Automated Learning of Features

We propose a novel high-performance, interpretable, and parameter \& computationally efficient deep learning architecture for tabular data, Gated Adaptive Network for Deep Automated Learning of Features (GANDALF). GANDALF relies on a new tabular processing unit with a gating mechanism and in-built feature selection called Gated Feature Learning Unit (GFLU) as a feature representation learning unit. We demonstrate that GANDALF outperforms or stays at-par with SOTA approaches like XGBoost, SAINT, FT-Transformers, etc. by experiments on multiple established public benchmarks. We have made available the code at github.com/manujosephv/pytorch_tabular under MIT License.

cs.LG↗

arXiv2021

Multi-Image Visual Question Answering

While a lot of work has been done on developing models to tackle the problem of Visual Question Answering, the ability of these models to relate the question to the image features still remain less explored. We present an empirical study of different feature extraction methods with different loss functions. We propose New dataset for the task of Visual Question Answering with multiple image inputs having only one ground truth, and benchmark our results on them. Our final model utilising Resnet + RCNN image features and Bert embeddings, inspired from stacked attention network gives 39% word accuracy and 99% image accuracy on CLEVER+TinyImagenet dataset.

cs.CV↗