Searcharxiv⌕ Search

arXiv subjects

Shubham Mehrotra

Publications and source records attributed to Shubham Mehrotra.

6 recordsLinked to original sources

AgentScript-Eval: Benchmarking LLMs and Agents for Code Generation in an Enterprise DSL

Domain-specific languages (DSLs) are widely used in enterprise software, giving a compact interface to complex tasks, from data access to infrastructure automation. Platforms increasingly extend them to agents, where a developer declares the topics, permitted actions, and guardrails of an agent rather than writing general-purpose code. Large language models (LLMs) write general-purpose code well, but that ability has been measured almost entirely on languages abundant in public pretraining data, and enterprise DSLs are neither abundant nor public. We introduce AgentScript-Eval, an execution-graded benchmark of 100 real tasks for AgentScript, the declarative DSL that Salesforce uses in Agentforce to build conversational agents. Tasks span authoring an agent from a specification, debugging a broken definition, and optimizing a working one. Evaluating five models zero-shot, we find that the regime is hard: the best model reaches only 25.0% pass@3, and every model scores 0% on authoring from scratch. We further find that model scale and price are weak predictors of success -- a 31B open-weight model attains the highest compilation rate of any model tested (48.0%) and matches the strongest frontier model on partial-credit accuracy at the lowest token cost. Adding an agentic compile-and-edit loop is the single largest lever, roughly doubling partial-credit accuracy. Beyond that gain, the wording of domain guidance matters more than its packaging. The same skill pack scores 28% as plain documentation but only 14% as a live plugin. Rewriting one skill inside that plugin then raises it to 37%, our best result, with catalog and tooling untouched. Larger models did not reliably outperform smaller ones; what mattered was the harness surrounding the model and the guidance supplied to it.

cs.SE↗

Koa-decision: Fast and Consistent Structured Decision Making with Generative LLMs

Industry applications often demand low-latency classification, yet current large language model (LLM) approaches remain poorly suited for latency-critical applications. Existing prompting and constrained decoding produce verbose, multi-token outputs that require expensive token-by-token generation, while encoder-based models achieve faster inference but sacrifice task flexibility. We propose Koa-decision, a framework for low-latency atomic actions -- fast, single-step decisions such as classification, semantic endpointing, Boolean checks, and scoring -- formulated as constrained generation with single-token outputs. By introducing atomic label tokens and applying supervised fine-tuning, our method reduces classification to a deterministic one-step decoding problem. Across standard benchmarks, Koa-decision delivers competitive accuracy with consistently low and stable latency. On a production intent-routing benchmark, Koa-decision reaches 85.5% accuracy -- competitive with the strongest frontier models (Claude-4.8-Opus, Gemini-Pro-3.1) and ahead of GPT-5 and Gemini-2.5-Pro -- while answering in about half a second, several-fold faster than every frontier model (up to ~7.5x at the median) under identical serving conditions. Against the dedicated single-token system Jev/TypeSafe, Koa-decision is competitive on accuracy and faster at the median, while also handling multimodal inputs and multi-label outputs that single-label text systems do not.

cs.LG↗

Salesforce Koa: An Enterprise Language Model for Agentic Tool Use

We present Salesforce Koa, an enterprise language model built by post-training the open-weight Nemotron-3-Super-120B foundation model with reinforcement learning using Group Relative Policy Optimization (GRPO), and deployed in FP8 for production. Koa is trained only on public and synthetically generated data, and specialized for the agentic tool use that enterprise workflows demand: routing a request to the correct action, invoking the right tool with valid arguments, and completing multi-turn business tasks. The distinctive component of our pipeline is specification-driven task construction: declarative Agent Script specifications are expanded into persona-conditioned multi-turn environments whose rewards are grounded in successful tool use. Applied to enterprise CRM specifications, the same pipeline produces the in-domain training distribution on which Koa is specialized. On CRMAgentBench and the human-labeled production tool-calling set, Koa outperforms both its untuned open-weight base and GPT-4.1 and is competitive with the strongest frontier models. It reaches 87% Task Success Rate on CRMAgentBench (vs. GPT-4.1 at 82% and the base at 79%), is at or near the top of every metric on the human-labeled portion of an internal production benchmark, and preserves the base model's general capability on public benchmarks (Tau2Bench, BFCL). A controlled comparison with architecture and RL recipe held fixed shows that the additional in-domain RL stage improves argument accuracy and full tool-call success on the human-labeled enterprise benchmark.

cs.CL↗

Co-Evolving Harnesses and Models: On-Policy Correction Helps Weaker Models Catch Up Where Imitation Fails

Agent harnesses (the system prompt, tool set, execution hooks, and context-management scaffolding around a model) are a critical determinant of agentic task success. Automated harness evolution can enable smaller models to perform well on domain-specific tasks at a fraction of frontier-model cost. Since both the harness and model weights shape behavior, we ask how harness evolution and lightweight fine-tuning should be combined. Across seven enterprise agent tasks, we first evolve a harness with the weaker model, then find that a stronger expert often uses it more effectively, suggesting expert supervision could close the remaining gap. However, training the weaker model on the expert's complete trajectories under the evolved harness backfires: performance regresses on all seven tasks by 4 to 30 points across Qwen3-Coder and Gemma 4, even though the same procedure helps under the unevolved harness. Our analysis shows that imitation transfers knowledge and increases scaffold usage, but disrupts model-harness fit: the weaker model adopts the expert's planning strategy without the competence to execute it and no longer matches the harness evolved around its native planning style. We therefore develop an on-policy expert-correction pipeline, automated by a meta-level MLE agent, that localizes the failing turn in the weaker model's own rollout and asks the expert to rewrite only that turn. This preserves the model's planning style and combines the gains of harness evolution and model adaptation. Our results identify and resolve a source of contention between harness and weight updates, yielding a compatibility-preserving recipe for economical co-evolution on domain-specific enterprise tasks.

cs.AI↗

Reinforcement Learning for LLM Post-Training: A Survey

Large language models (LLMs) trained via pretraining and supervised fine-tuning (SFT) can still produce harmful and misaligned outputs, or struggle in domains like math and coding. Reinforcement learning (RL)-based post-training methods, including Reinforcement Learning from Human Feedback (RLHF) methods like Direct Preference Optimization (DPO) and Reinforcement Learning with Verifiable Rewards (RLVR) approaches like PPO and GRPO, have made remarkable gains to alleviate these issues. Yet, no existing work offers a technically detailed comparison of the various methods driving this progress. In order to fill this gap, we present a timely survey that connects foundational components with latest advancements. We derive a single policy gradient framework that unifies pretraining, SFT, RLHF, and RLVR as special cases while also organizing the more recent techniques therein. The main contributions of our survey are as follows: (1) a self-contained introduction to MLE, RLHF, and RLVR foundations and the unified policy gradient framework; (2) detailed technical analysis of PPO- and GRPO-based methods alongside offline and iterative DPO approaches, decomposed along prompt sampling, response sampling, and gradient coefficient axes; (3) standardized notation enabling direct cross-method comparison; and (4) comprehensive comparison of implementation details and empirical results of each method in the appendix. We aim to serve as a technically grounded reference for researchers and practitioners working on LLM post-training.

cs.CL↗

Rate, Explain and Cite (REC): Enhanced Explanation and Attribution in Automatic Evaluation by Large Language Models

LLMs have demonstrated impressive proficiency in generating coherent and high-quality text, making them valuable across a range of text-generation tasks. However, rigorous evaluation of this generated content is crucial, as ensuring its quality remains a significant challenge due to persistent issues such as factual inaccuracies and hallucination. This paper introduces three fine-tuned general-purpose LLM autoevaluators, REC-8B, REC-12B and REC-70B, specifically designed to evaluate generated text across several dimensions: faithfulness, instruction following, coherence, and completeness. These models not only provide ratings for these metrics but also offer detailed explanation and verifiable citation, thereby enhancing trust in the content. Moreover, the models support various citation modes, accommodating different requirements for latency and granularity. Extensive evaluations on diverse benchmarks demonstrate that our general-purpose LLM auto-evaluator, REC-70B, outperforms state-of-the-art LLMs, excelling in content evaluation by delivering better quality explanation and citation with minimal bias. Our REC dataset and models are available at https://github.com/adelaidehsu/REC.

cs.CL↗