Training-time privileged information (PI) can enable language models to succeed on tasks they would otherwise fail, making it a powerful tool for reinforcement learning in hard, long-horizon settings. However, transferring capabilities learned with PI to policies that must act without it at inference time remains a fundamental challenge. We study this problem in the context of distilling frontier models for multi-turn agentic environments, where closed-source systems typically hide their internal reasoning and expose only action trajectories. This breaks standard distillation pipelines, since successful behavior is observable but the reasoning process is not. We introduce π-Distill, a joint teacher–student framework that trains a PI-conditioned teacher and an unconditioned student simultaneously within a single shared-parameter model, enabling the teacher to learn how to use PI while mitigating distribution shift during transfer. We show that π-Distill effectively distills frontier agents using action-only privileged information, matching or outperforming industry-standard pipelines that assume access to full Chain-of-Thought supervision across multiple agentic benchmarks, models, and forms of PI. We complement our results with extensive analysis that characterize what factors enable effective learning with PI.
Voice agents, artificial intelligence systems that conduct spoken conversations to complete tasks, are increasingly deployed across enterprise applications. However, no existing benchmark jointly addresses two core evaluation challenges: generating realistic simulated conversations, and measuring quality across the full scope of voice-specific failure modes. We present EVA-Bench, an end-to-end evaluation framework that addresses both. On the simulation side, EVA-Bench orchestrates bot-to-bot audio conversations over dynamic multi-turn dialogues, with automatic simulation validation that detects user simulator error and appropriately regenerates conversations before scoring. On the measurement side, EVA-Bench introduces two composite metrics: EVA-A (Accuracy), capturing task completion, faithfulness, and audio-level speech fidelity; and EVA-X (Experience), capturing conversation progression, spoken conciseness, and turn-taking timing. Both metrics apply to different agent architectures, enabling direct cross-architecture comparison. EVA-Bench includes 213 scenarios across three enterprise domains, a controlled perturbation suite for accent and noise robustness, and pass@1, pass@k, pass^k measurements that distinguish peak from reliable capability. Across 12 systems spanning all three architectures, we find: (1) no system simultaneously exceeds 0.5 on both EVA-A pass@1 and EVA-X pass@1; (2) peak and reliable performance diverge substantially (median pass@k - pass^k gap of 0.44 on EVA-A); and (3) accent and noise perturbations expose substantial robustness gaps, with effects varying across architectures, systems, and metrics (mean up to 0.314). We release the full framework, evaluation suite, and benchmark data under an open-source license.
Test-time scaling has significantly improved large language model (LLM) performance, enabling deeper reasoning to solve complex problems. However, this increased reasoning capability also leads to excessive token generation and unnecessary problem-solving attempts. We introduce "Don't Reason Bench (DNR Bench)", a new benchmark designed to evaluate LLMs’ ability to robustly understand tricky reasoning triggers and avoid unnecessary generation. DNR Bench consists of 150 adversarially designed prompts that are easy for humans to understand and respond to, but surprisingly not for many recent prominent LLMs. DNR Bench tests models' abilities across different capabilities, such as instruction adherence, hallucination avoidance, redundancy filtering, and unanswerable question recognition. We evaluate reasoning LLMs (RLMs), including DeepSeek-R1, OpenAI O3-mini, and Claude-3.7-sonnet, and compare them against a powerful non-reasoning model, such as GPT-4o. Our experiments reveal that RLMs generate up to 70x more tokens than necessary, often failing at tasks that simpler non-reasoning models handle efficiently with higher accuracy. Our findings underscore the need for more effective training and inference strategies in RLMs.
Enterprise LLM applications require consistently high quality and reliable performance across diverse scenarios, demanding robustness to minor variations. Existing research shows that even small prompt changes can lead to substantial differences in output, but has mainly focused on a narrow set of perturbations with small academic datasets, limiting their relevance to real-world applications. To address this, we present a comprehensive benchmark suite that evaluates robustness across multiple perturbation types, including general text edits (e.g., punctuation, whitespace), formatting changes (e.g., JSON, YAML), multilingual and cross-lingual inputs, and positional variations in instructions. Evaluating 11 models ranging from 4B to 120B+ parameters, we find that minor perturbations reduce performance by up to 40 percentage points on key enterprise metrics. Critically, we demonstrate that the relationship between model size and robustness is more nuanced than conventional assumptions suggest: an 8B parameter model (Ministral 3 8B) outperforms most larger models, while another 8B model (Llama 3.1 8B) performs worst overall.
Workflows are a fundamental component of automation in enterprise platforms, enabling the orchestration of tasks, data processing, and system integrations. Despite being widely used, building workflows can be complex, often requiring manual configuration through low-code platforms or visual programming tools. To simplify this process, we explore the use of generative foundation models, particularly vision-language models (VLMs), to automatically generate structured workflows from visual inputs. Translating hand-drawn sketches or computer-generated diagrams into executable workflows is challenging due to the ambiguity of free-form drawings, variations in diagram styles, and the difficulty of inferring execution logic from visual elements. To address this, we introduce StarFlow, a framework for generating structured workflow outputs from sketches using vision-language models. We curate a diverse dataset of workflow diagrams – including synthetic, manually annotated, and real-world samples – to enable robust training and evaluation. We finetune and benchmark multiple vision-language models, conducting a series of ablation studies to analyze the strengths and limitations of our approach. Our results show that finetuning significantly enhances structured workflow generation, outperforming large vision-language models on this task.
