Research

Universal deep research agent team. 13-agent pipeline for rigorous academic research on any topic. 7 modes: full research, quick brief, paper review, lit-review, fact-check, Socratic guided research dialogue, and systematic review with optional meta-analysis. Covers research question formulation, Socratic mentoring, methodology design, systematic literature search, source verification, cross-source synthesis, risk of bias assessment, meta-analysis, APA 7.0 report compilation, editorial review...

Orchestrator for the full academic research pipeline: research -> write -> integrity check -> review -> revise -> re-review -> re-revise -> final integrity check -> finalize. Coordinates deep-research, academic-paper, and academic-paper-reviewer into a seamless 10-stage workflow with mandatory integrity verification, two-stage peer review, and reproducible quality gates. Triggers on: academic pipeline, research to paper, full paper workflow, paper pipeline, end-to-end paper, research-to-publi...

12-agent academic paper writing pipeline. 10 modes (full/plan/outline/revision/revision-coach/abstract/lit-review/format-convert/citation-check/disclosure). 6 paper types, 5 citation formats, bilingual abstracts, LaTeX/DOCX-via-Pandoc/PDF output. Style Calibration + Writing Quality Check + Anti-Patterns with IRON RULE markers. Triggers: write paper, academic paper, guide my paper, parse reviews, AI disclosure, 寫論文, 學術論文, 引導我寫論文, 審查意見.

Multi-perspective academic paper review with dynamic reviewer personas. Simulates 5 independent reviewers (EIC + 3 peer reviewers + Devil's Advocate) with field-specific expertise. Supports full review, re-review (verification), quick assessment, methodology focus, Socratic guided, and calibration modes. Triggers on: review paper, peer review, manuscript review, referee report, review my paper, critique paper, simulate review, editorial review, calibrate reviewer, reviewer calibration, measur...

Multi-agent automated scientific discovery for diseases — given a disease name, Robin generates and ranks experimental assays, proposes therapeutic candidates, and (optionally) analyzes wet-lab data. Open-source, Apache-2.0. Use when the user wants an end-to-end "I have a disease, give me hypotheses to test" workflow rather than a single literature lookup.

Generate Wikipedia-style scientific articles section-by-section by orchestrating PaperQA2 over a topic-specific corpus. Reproduces the WikiCrow recipe used by FutureHouse to write the gene articles at wikicrow.ai. Use when the user wants a structured, fully-cited long-form article on a scientific topic (gene, protein, disease, drug, mechanism) rather than a single Q&A answer.

"Has anyone done this before?" — precedent search across the scientific literature via FutureHouse's Owl agent (formerly HasAnyone). Use when the user wants to know whether a specific experiment, technique, measurement, drug-target combination, or method has ever been published. Returns a yes/no-grounded answer with the closest matching prior work.

Deep, high-reasoning literature synthesis via FutureHouse's Falcon agent (LITERATURE_HIGH job). Use when the user wants a thematic review, gap analysis, or systematic synthesis across many papers — not a single fact lookup. Costs more credits and takes minutes longer than Crow but produces SOTA-quality scholarly output.

Fast scientific literature Q&A with citations via FutureHouse's Crow agent (production PaperQA2). Use when the user wants a single, well-cited answer drawn from the published scientific literature — biology, chemistry, medicine, ML, etc. Handles one focused question per call. For multi-paper thematic synthesis use Falcon; for "has anyone done X" precedent queries use Owl.

Compute pass@k — the standard "any of N samples is correct" metric for code-generation evaluation (HumanEval / MBPP / LiveCodeBench / APPS / BigCodeBench / CodeContests). Use when the user has N samples per problem and wants the unbiased estimator of "probability at least one of the top-k is correct". Returns mean pass@k across the dataset, in [0, 1].

Compute normalized Discounted Cumulative Gain at cutoff k (nDCG@k) — the standard ranking metric for retrieval / recommendation / search evaluation when relevance is graded. Use when the user has a list of (query, ranked_doc_ids, relevance_judgements) and wants to score the ranking quality, or mentions "nDCG / NDCG / DCG / ranking metric / IR metric / BEIR-style eval". Returns a number in [0, 1]; higher = better ranking.

Evaluate an LLM on GSM8K — 1K grade-school math word problems requiring 2-8 step arithmetic reasoning. Use when the user wants to measure math reasoning, mentions GSM8K, or asks "how good is my model at multi-step word problems?". Reports exact-match accuracy on the final numeric answer (parsed from "#### N" suffix).

Evaluates zero-shot cross-lingual news recommendation by measuring how effectively a model recommends articles in a target language to users who only consume news in a source language. It probes the model's ability to leverage multilingual sentence embeddings and click behavior fusion without task-specific fine-tuning on the target language. Use when the user wants to benchmark on MIND (small) / xMIND (small), or asks about evaluating this task. Reports nDCG@10.

Evaluates zero-shot multimodal named entity typing and relation extraction. It probes a model's ability to align text and image modalities for fine-grained semantic recognition of unseen entity types and relations without additional training. Use when the user wants to benchmark on WikiDiverse, Zheng et al. MRE dataset, or asks about evaluating this task. Reports F1.

Evaluates a model's ability to retrieve target images based on a reference image and a natural language modification text. It probes fine-grained visual-semantic alignment, compositional reasoning, and ranking precision under varying levels of distractors and semantic transformations. Use when the user wants to benchmark on CIRR, CIRCO, FashionIQ, GeneCIS, or asks about evaluating this task. Reports Recall@K.

Evaluates fine-grained multimodal perception, visual grounding, and reasoning capabilities of vision-language models. It measures performance across general perception, specific perception (color, counting), and out-of-distribution generalization tasks using a suite of established and custom benchmarks. Use when the user wants to benchmark on ZoomBench, HR-Bench, VStar, CV-Bench, MME-RealWorld, ColorBench, CountQA, MMStar, BabyVision, or asks about evaluating this task. Reports accuracy.

This benchmark evaluates non-autoregressive spoken dialogue generation models on their ability to produce multi-turn conversational audio that matches input text, maintains speaker identity, and accurately handles turn-taking between two speakers. It probes both objective speech quality metrics and subjective human judgments of coherence and similarity. Use when the user wants to benchmark on test-dialog-zh, test-dialog-en, or asks about evaluating this task. Reports cpWER.

Evaluates de novo molecular generation models on the ZINC-250k dataset across three tasks: unconditional generation from a Gaussian prior, unconstrained property optimization (maximizing penalized logP and QED), and constrained optimization (modifying molecules to improve logP while preserving structural similarity). Use when the user wants to benchmark on ZINC-250k, or asks about evaluating this task. Reports Top-3 Penalized logP.

Evaluates a model's ability to understand and generalize across unseen NLP tasks based solely on task descriptions, rather than few-shot examples. It probes systematic generalization across variations like paraphrasing, composition, semantic flips, and output structure changes. Use when the user wants to benchmark on ZEST, or asks about evaluating this task. Reports Mean.

Evaluates a model's ability to perform visual-text compression (OCR) by measuring raw text retention on rendered documents where the textual content has been deliberately stripped of semantic meaning. It isolates pure visual decoding capability from downstream linguistic priors or contextual inference. Use when the user wants to benchmark on ZeroSense, or asks about evaluating this task. Reports text preservation capability.

Evaluates the accuracy and inference latency of post-training quantized Transformer models (BERT and GPT-3-style) on standard NLP benchmarks and language modeling tasks. Use when the user wants to benchmark on GLUE benchmark, 20 zero-shot evaluation tasks, PTB / Wikitext-2 / Wikitext-103, or asks about evaluating this task. Reports average accuracy.

Evaluates zero-shot speech synthesis and novel voice generation by measuring how well models can produce intelligible, natural, and speaker-similar audio for unseen speakers using only conditioning embeddings. Use when the user wants to benchmark on English Multi-Accent Dataset (VCTK + Internal), or asks about evaluating this task. Reports WER.

Evaluates zero-shot text-to-speech models on Vietnamese speech generation, measuring intelligibility, speaker similarity, and naturalness across long-form and short-form text inputs. Use when the user wants to benchmark on viVoice, PAB-S, PAB-U, VIVOS, or asks about evaluating this task. Reports WER.

Evaluates zero-shot text-to-speech synthesis capability across English and Chinese. It measures intelligibility, speaker similarity, and naturalness against reference prompts without fine-tuning on target speakers. Use when the user wants to benchmark on Seed-TTS test-en, Seed-TTS test-zh, AISHELL-3 test set, or asks about evaluating this task. Reports WER.