| import React from "react"; | |
| import { | |
| Box, | |
| Typography, | |
| Paper, | |
| IconButton, | |
| Tooltip, | |
| Alert, | |
| Link, | |
| } from "@mui/material"; | |
| import ContentCopyIcon from "@mui/icons-material/ContentCopy"; | |
| import PageHeader from "../../components/shared/PageHeader"; | |
| const citations = [ | |
| { | |
| title: "LLM Security Leaderboard", | |
| authors: | |
| "Pankaj Telang, Eleftheria Stein-Kousathana, Doug Wright, Daniel Kantor, Luke Hinds", | |
| citation: `@misc{llm-security-leaderboard-v1, | |
| author = {Pankaj Telang and Eleftheria Stein-Kousathana and Doug Wright and Daniel Kantor and Luke Hinds}, | |
| title = {LLM Security Leaderboard}, | |
| year = {2025}, | |
| publisher = {Hugging Face}, | |
| howpublished = "\\url{https://huggingface.co/spaces/stacklok/llm-security-leaderboard}", | |
| }`, | |
| type: "main", | |
| }, | |
| { | |
| title: "LLM Security Leaderboard Evaluation Backend", | |
| authors: | |
| "Pankaj Telang, Eleftheria Stein-Kousathana, Doug Wright, Daniel Kantor, Luke Hinds", | |
| citation: `@misc{stacklok/llm-security-evaluation-backend-v1, | |
| author = {Pankaj Telang and Eleftheria Stein-Kousathana and Doug Wright and Daniel Kantor and Luke Hinds}, | |
| title = {LLM Security Leaderboard Evaluation Backend}, | |
| year = {2025}, | |
| publisher = {Hugging Face}, | |
| howpublished = "\\url{https://huggingface.co/spaces/stacklok/llm-security-evaluation-backend}", | |
| }`, | |
| type: "main", | |
| }, | |
| ]; | |
| const priorWork = [ | |
| { | |
| title: "Open LLM Leaderboard v2", | |
| authors: | |
| "Clémentine Fourrier, Nathan Habib, Alina Lozovskaya, Konrad Szafer, Thomas Wolf", | |
| citation: `@misc{open-llm-leaderboard-v2, | |
| author = {Clémentine Fourrier and Nathan Habib and Alina Lozovskaya and Konrad Szafer and Thomas Wolf}, | |
| title = {Open LLM Leaderboard v2}, | |
| year = {2024}, | |
| publisher = {Hugging Face}, | |
| howpublished = "\\url{https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard}", | |
| }`, | |
| type: "main", | |
| }, | |
| { | |
| title: "Open LLM Leaderboard v1", | |
| authors: | |
| "Edward Beeching, Clémentine Fourrier, Nathan Habib, Sheon Han, Nathan Lambert, Nazneen Rajani, Omar Sanseviero, Lewis Tunstall, Thomas Wolf", | |
| citation: `@misc{open-llm-leaderboard-v1, | |
| author = {Edward Beeching and Clémentine Fourrier and Nathan Habib and Sheon Han and Nathan Lambert and Nazneen Rajani and Omar Sanseviero and Lewis Tunstall and Thomas Wolf}, | |
| title = {Open LLM Leaderboard (2023-2024)}, | |
| year = {2023}, | |
| publisher = {Hugging Face}, | |
| howpublished = "\\url{https://huggingface.co/spaces/open-llm-leaderboard-old/open_llm_leaderboard}" | |
| }`, | |
| type: "main", | |
| }, | |
| { | |
| "title": "CYBERSECEVAL 3: Advancing the Evaluation of Cybersecurity Risks and Capabilities in Large Language Models", | |
| "authors": "Shengye Wan, Cyrus Nikolaidis, Daniel Song, David Molnar, James Crnkovich, Jayson Grace, Manish Bhatt, Sahana Chennabasappa, Spencer Whitman, Stephanie Ding, Vlad Ionescu, Yue Li, Joshua Saxe", | |
| "citation": `@misc{wan2024cyberseceval3, | |
| author = {Shengye Wan and Cyrus Nikolaidis and Daniel Song and David Molnar and James Crnkovich and Jayson Grace and Manish Bhatt and Sahana Chennabasappa and Spencer Whitman and Stephanie Ding and Vlad Ionescu and Yue Li and Joshua Saxe}, | |
| title = {{CYBERSECEVAL 3}: Advancing the Evaluation of Cybersecurity Risks and Capabilities in Large Language Models}, | |
| year = {2024}, | |
| archivePrefix = {arXiv}, | |
| eprint = {2408.01605}, | |
| primaryClass = {cs.CR}, | |
| url = {https://doi.org/10.48550/arXiv.2408.01605} | |
| }`, | |
| "type": "main" | |
| } | |
| ]; | |
| const benchmarks = [ | |
| { | |
| title: "Bad Package Detection", | |
| authors: | |
| "Pankaj Telang, Eleftheria Stein-Kousathana, Doug Wright, Daniel Kantor, Luke Hinds", | |
| citation: `@misc{llm-security-leaderboard-v1-bad-package-v1, | |
| author = {Pankaj Telang and Eleftheria Stein-Kousathana and Doug Wright and Daniel Kantor and Luke Hinds}, | |
| title = {Bad Package Detection}, | |
| year = {2025}, | |
| publisher = {Hugging Face}, | |
| howpublished = "\\url{https://huggingface.co/spaces/stacklok/llm-security-leaderboard}", | |
| }`, | |
| type: "main", | |
| }, | |
| { | |
| title: "Insecure Code Detection", | |
| authors: | |
| "Pankaj Telang, Eleftheria Stein-Kousathana, Doug Wright, Daniel Kantor, Luke Hinds", | |
| citation: `@misc{llm-security-leaderboard-v1-insecure-code-v1, | |
| author = {Pankaj Telang and Eleftheria Stein-Kousathana and Doug Wright and Daniel Kantor and Luke Hinds}, | |
| title = {Insecure Code Detection}, | |
| year = {2025}, | |
| publisher = {Hugging Face}, | |
| howpublished = "\\url{https://huggingface.co/spaces/stacklok/llm-security-leaderboard}", | |
| }`, | |
| type: "main", | |
| }, | |
| { | |
| title: "Safetensors Usage", | |
| authors: | |
| "Pankaj Telang, Eleftheria Stein-Kousathana, Doug Wright, Daniel Kantor, Luke Hinds", | |
| citation: `@misc{llm-security-leaderboard-v1-safetensors-v1, | |
| author = {Pankaj Telang and Eleftheria Stein-Kousathana and Doug Wright and Daniel Kantor and Luke Hinds}, | |
| title = {Safetensors Usage}, | |
| year = {2025}, | |
| publisher = {Hugging Face}, | |
| howpublished = "\\url{https://huggingface.co/spaces/stacklok/llm-security-leaderboard}", | |
| }`, | |
| type: "main", | |
| }, | |
| { | |
| title: "CVE Knowledge", | |
| authors: | |
| "Pankaj Telang, Eleftheria Stein-Kousathana, Doug Wright, Daniel Kantor, Luke Hinds", | |
| citation: `@misc{llm-security-leaderboard-v1-cve-v1, | |
| author = {Pankaj Telang and Eleftheria Stein-Kousathana and Doug Wright and Daniel Kantor and Luke Hinds}, | |
| title = {CVE Knowledge}, | |
| year = {2025}, | |
| publisher = {Hugging Face}, | |
| howpublished = "\\url{https://huggingface.co/spaces/stacklok/llm-security-leaderboard}", | |
| }`, | |
| type: "main", | |
| }, | |
| ]; | |
| const CitationBlock = ({ citation, title, authors, url, type }) => { | |
| const handleCopy = () => { | |
| navigator.clipboard.writeText(citation); | |
| }; | |
| return ( | |
| <Paper | |
| elevation={0} | |
| sx={{ | |
| p: 3, | |
| border: "1px solid", | |
| borderColor: "grey.200", | |
| backgroundColor: "transparent", | |
| borderRadius: 2, | |
| position: "relative", | |
| }} | |
| > | |
| <Box sx={{ mb: 2 }}> | |
| <Typography variant="h6" sx={{ mb: 0.5 }}> | |
| {title} | |
| </Typography> | |
| <Typography variant="body2" color="text.secondary"> | |
| {authors} | |
| </Typography> | |
| {url && ( | |
| <Link | |
| href={url} | |
| target="_blank" | |
| rel="noopener noreferrer" | |
| sx={{ fontSize: "0.875rem", display: "block", mt: 0.5 }} | |
| > | |
| View paper → | |
| </Link> | |
| )} | |
| </Box> | |
| <Box | |
| sx={{ | |
| backgroundColor: "grey.900", | |
| borderRadius: 1, | |
| p: 2, | |
| position: "relative", | |
| }} | |
| > | |
| <Tooltip title="Copy citation" placement="top"> | |
| <IconButton | |
| onClick={handleCopy} | |
| size="small" | |
| sx={{ | |
| position: "absolute", | |
| top: 8, | |
| right: 8, | |
| color: "grey.500", | |
| "&:hover": { | |
| color: "grey.300", | |
| }, | |
| }} | |
| > | |
| <ContentCopyIcon fontSize="small" /> | |
| </IconButton> | |
| </Tooltip> | |
| <Box | |
| component="pre" | |
| sx={{ | |
| margin: 0, | |
| color: "#fff", | |
| fontSize: "0.875rem", | |
| fontFamily: "monospace", | |
| whiteSpace: "pre", | |
| textAlign: "left", | |
| overflow: "auto", | |
| }} | |
| > | |
| <code>{citation}</code> | |
| </Box> | |
| </Box> | |
| </Paper> | |
| ); | |
| }; | |
| function QuotePage() { | |
| return ( | |
| <Box sx={{ width: "100%", maxWidth: 1200, margin: "0 auto", py: 4, px: 0 }}> | |
| <PageHeader | |
| title="Citation Information" | |
| subtitle="How to cite the Open LLM Leaderboard in your work" | |
| /> | |
| <Alert severity="info" sx={{ mb: 4 }}> | |
| <Typography variant="body2"> | |
| The citations below include both the leaderboard itself and the | |
| individual benchmarks used in our evaluation suite. | |
| </Typography> | |
| </Alert> | |
| <Box sx={{ mb: 6 }}> | |
| <Typography variant="h5" sx={{ mb: 3 }}> | |
| Leaderboard | |
| </Typography> | |
| <Box sx={{ display: "flex", flexDirection: "column", gap: 3 }}> | |
| {citations.map((citation, index) => ( | |
| <CitationBlock key={index} {...citation} /> | |
| ))} | |
| </Box> | |
| </Box> | |
| <Box sx={{ mb: 6 }}> | |
| <Typography variant="h5" sx={{ mb: 3 }}> | |
| Benchmarks | |
| </Typography> | |
| <Box sx={{ display: "flex", flexDirection: "column", gap: 3 }}> | |
| {benchmarks.map((benchmark, index) => ( | |
| <CitationBlock key={index} {...benchmark} /> | |
| ))} | |
| </Box> | |
| </Box> | |
| <Box> | |
| <Typography variant="h5" sx={{ mb: 3 }}> | |
| Prior Work | |
| </Typography> | |
| <Box sx={{ display: "flex", flexDirection: "column", gap: 3 }}> | |
| {priorWork.map((citation, index) => ( | |
| <CitationBlock key={index} {...citation} /> | |
| ))} | |
| </Box> | |
| </Box> | |
| </Box> | |
| ); | |
| } | |
| export default QuotePage; | |