Spaces:
Running
Running
| <div> <script type="text/javascript">window.PlotlyConfig = {MathJaxConfig: 'local'};</script> | |
| <div id="3dcf3570-b639-4c3c-94e9-d775eb388370" class="plotly-graph-div" style="height:100%; width:100%;"></div> <script type="text/javascript"> window.PLOTLYENV=window.PLOTLYENV || {}; if (document.getElementById("3dcf3570-b639-4c3c-94e9-d775eb388370")) { Plotly.newPlot( "3dcf3570-b639-4c3c-94e9-d775eb388370", [{"customdata":[["ARC",50.43,"bigscience\u002fbloom"],["ARC",70.48,"migtissera\u002fSynthia-70B-v1.2"],["ARC",72.1,"uni-tianyan\u002fUni-TianYan"],["ARC",72.27,"fangloveskari\u002fORCA_LLaMA_70B_QLoRA"],["ARC",73.12,"Riiid\u002fsheep-duck-llama-2-70b-v1.1"],["ARC",73.38,"fblgit\u002fUNAversal-2x7B-v1"],["ARC",73.55,"TomGrc\u002fFusionNet_7Bx2_MoE_14B"],["ARC",73.72,"Kquant03\u002fProkaryote-8x7B-bf16"],["ARC",74.91,"yunconglong\u002fTruthful_DPO_TomGrc_FusionNet_7Bx2_MoE_13B"],["ARC",76.02,"abacusai\u002fSmaug-72B-v0.1"],["ARC",77.47,"saltlux\u002fluxia-21.4b-alignment-v1.0"],["ARC",78.24,"4season\u002falignment_model_test"],["ARC",79.78,"davidkim205\u002fRhea-72b-v0.5"],["ARC",79.86,"freewheelin\u002ffree-evo-qwen72b-v0.8-re"]],"hovertemplate":"Model Name: %{customdata[2]}\u003cbr\u003eMetric Name: %{customdata[0]}\u003cbr\u003eDate: %{x}\u003cbr\u003eMetric Value: %{y}","legendgroup":"ARC","line":{"color":"#636efa","dash":"solid"},"marker":{"symbol":"circle"},"mode":"lines+markers","name":"ARC","orientation":"v","showlegend":true,"x":["2023-08-25T13:05:50+00:00","2023-09-09T10:52:17+00:00","2023-10-16T12:46:18+00:00","2023-10-16T12:48:18+00:00","2023-12-04T03:08:34+00:00","2024-01-09T08:46:56+00:00","2024-01-12T16:05:44+00:00","2024-01-18T04:27:19+00:00","2024-01-21T09:10:58+00:00","2024-02-03T18:49:25+00:00","2024-03-11T03:09:26+00:00","2024-03-16T11:32:40+00:00","2024-03-22T15:04:33+00:00","2024-05-05T07:26:59+00:00"],"xaxis":"x","y":[50.43,70.48,72.1,72.27,73.12,73.38,73.55,73.72,74.91,76.02,77.47,78.24,79.78,79.86],"yaxis":"y","type":"scatter"},{"customdata":[["MMLU",30.85,"bigscience\u002fbloom"],["MMLU",70.67,"jondurbin\u002fairoboros-l2-70b-gpt4-m2.0"],["MMLU",70.7,"ICBU-NPU\u002fFashionGPT-70B-V1"],["MMLU",70.99,"ICBU-NPU\u002fFashionGPT-70B-V1.1"],["MMLU",78.13,"bhenrym14\u002fplatypus-yi-34b"],["MMLU",78.46,"kyujinpy\u002fPlatYi-34B-LoRA"],["MMLU",78.78,"kyujinpy\u002fPlatYi-34B-Llama-Q"],["MMLU",85.6,"CausalLM\u002f34b-beta"]],"hovertemplate":"Model Name: %{customdata[2]}\u003cbr\u003eMetric Name: %{customdata[0]}\u003cbr\u003eDate: %{x}\u003cbr\u003eMetric Value: %{y}","legendgroup":"MMLU","line":{"color":"#EF553B","dash":"solid"},"marker":{"symbol":"circle"},"mode":"lines+markers","name":"MMLU","orientation":"v","showlegend":true,"x":["2023-08-25T13:05:50+00:00","2023-09-09T10:52:17+00:00","2023-10-16T13:19:55+00:00","2023-11-06T10:31:15+00:00","2023-11-15T16:49:09+00:00","2023-12-03T12:09:22+00:00","2023-12-06T11:25:55+00:00","2024-02-09T16:11:28+00:00"],"xaxis":"x","y":[30.85,70.67,70.7,70.99,78.13,78.46,78.78,85.6],"yaxis":"y","type":"scatter"},{"customdata":[["Winogrande",72.06,"bigscience\u002fbloom"],["Winogrande",84.29,"augtoma\u002fqCammel-70-x"],["Winogrande",86.9,"tiiuae\u002ffalcon-180B"],["Winogrande",88.16,"TomGrc\u002fFusionNet_7Bx2_MoE_14B"],["Winogrande",88.24,"yunconglong\u002fTruthful_DPO_TomGrc_FusionNet_7Bx2_MoE_13B"],["Winogrande",88.4,"RubielLabarta\u002fLogoS-7Bx2-MoE-13B-v0.2"]],"hovertemplate":"Model Name: %{customdata[2]}\u003cbr\u003eMetric Name: %{customdata[0]}\u003cbr\u003eDate: %{x}\u003cbr\u003eMetric Value: %{y}","legendgroup":"Winogrande","line":{"color":"#00cc96","dash":"solid"},"marker":{"symbol":"circle"},"mode":"lines+markers","name":"Winogrande","orientation":"v","showlegend":true,"x":["2023-08-25T13:05:50+00:00","2023-09-09T10:52:17+00:00","2023-09-15T11:25:37.918274+00:00","2024-01-12T16:05:44+00:00","2024-01-21T09:10:58+00:00","2024-02-11T16:05:38+00:00"],"xaxis":"x","y":[72.06,84.29,86.9,88.16,88.24,88.4],"yaxis":"y","type":"scatter"},{"customdata":[["HellaSwag",76.41,"bigscience\u002fbloom"],["HellaSwag",88.32,"TheBloke\u002fllama-2-70b-Guanaco-QLoRA-fp16"],["HellaSwag",88.89,"tiiuae\u002ffalcon-180B"],["HellaSwag",90.11,"Sao10K\u002fWinterGoddess-1.4x-70B-L2"],["HellaSwag",91.88,"saltlux\u002fluxia-21.4b-alignment-v1.0"],["HellaSwag",92.75,"swap-uniba\u002fLLaMAntino-3-ANITA-8B-Inst-DPO-ITA"]],"hovertemplate":"Model Name: %{customdata[2]}\u003cbr\u003eMetric Name: %{customdata[0]}\u003cbr\u003eDate: %{x}\u003cbr\u003eMetric Value: %{y}","legendgroup":"HellaSwag","line":{"color":"#ab63fa","dash":"solid"},"marker":{"symbol":"circle"},"mode":"lines+markers","name":"HellaSwag","orientation":"v","showlegend":true,"x":["2023-08-25T13:05:50+00:00","2023-09-09T10:52:17+00:00","2023-09-15T11:25:37.918274+00:00","2023-12-14T09:05:41+00:00","2024-03-11T03:09:26+00:00","2024-05-14T11:29:36+00:00"],"xaxis":"x","y":[76.41,88.32,88.89,90.11,91.88,92.75],"yaxis":"y","type":"scatter"},{"customdata":[["GSM8K",6.9,"bigscience\u002fbloom"],["GSM8K",35.33,"OpenLemur\u002flemur-70b-chat-v1"],["GSM8K",45.94,"tiiuae\u002ffalcon-180B"],["GSM8K",56.41,"abhishek\u002fzephyr-beta-math"],["GSM8K",59.82,"bhenrym14\u002fplatypus-yi-34b"],["GSM8K",60.8,"chargoddard\u002fYi-34B-Llama"],["GSM8K",62.7,"CausalLM\u002f14B-DPO-alpha"],["GSM8K",63.68,"deepseek-ai\u002fdeepseek-llm-67b-chat"],["GSM8K",68.84,"meta-math\u002fMetaMath-Mistral-7B"],["GSM8K",71.49,"Q-bert\u002fMetaMath-Cybertron-Starling"],["GSM8K",72.18,"SUSTech\u002fSUS-Chat-34B"],["GSM8K",73.16,"Toten5\u002fMarcoroni-neural-chat-7B-v2"],["GSM8K",73.31,"samir-fama\u002fFernandoGPT-v1"],["GSM8K",76.27,"moreh\u002fMoMo-72B-lora-1.8.4-DPO"],["GSM8K",76.8,"moreh\u002fMoMo-72B-lora-1.8.6-DPO"],["GSM8K",78.62,"moreh\u002fMoMo-72B-lora-1.8.7-DPO"],["GSM8K",78.7,"abacusai\u002fSmaug-72B-v0.1"],["GSM8K",82.03,"mistralai\u002fMixtral-8x22B-Instruct-v0.1"],["GSM8K",85.44,"meta-llama\u002fMeta-Llama-3-70B-Instruct"],["GSM8K",86.05,"MaziyarPanahi\u002fLlama-3-70B-Instruct-DPO-v0.1"],["GSM8K",86.28,"tenyx\u002fLlama3-TenyxChat-70B"],["GSM8K",88.25,"MaziyarPanahi\u002fLlama-3-70B-Instruct-DPO-v0.2"]],"hovertemplate":"Model Name: %{customdata[2]}\u003cbr\u003eMetric Name: %{customdata[0]}\u003cbr\u003eDate: %{x}\u003cbr\u003eMetric Value: %{y}","legendgroup":"GSM8K","line":{"color":"#FFA15A","dash":"solid"},"marker":{"symbol":"circle"},"mode":"lines+markers","name":"GSM8K","orientation":"v","showlegend":true,"x":["2023-08-25T13:05:50+00:00","2023-09-09T10:52:17+00:00","2023-09-15T11:25:37.918274+00:00","2023-11-14T13:30:22+00:00","2023-11-15T16:49:09+00:00","2023-11-17T13:40:31+00:00","2023-11-19T20:55:46+00:00","2023-12-01T07:45:30+00:00","2023-12-03T15:31:15+00:00","2023-12-05T20:10:40+00:00","2023-12-09T04:51:31+00:00","2023-12-12T16:18:51+00:00","2023-12-31T09:50:37+00:00","2024-01-12T06:33:27+00:00","2024-01-16T11:52:34+00:00","2024-01-22T00:16:35+00:00","2024-02-03T18:49:25+00:00","2024-04-17T15:30:22+00:00","2024-04-18T17:05:16+00:00","2024-04-26T19:35:47+00:00","2024-04-27T17:40:55+00:00","2024-05-01T17:04:08+00:00"],"xaxis":"x","y":[6.9,35.33,45.94,56.41,59.82,60.8,62.7,63.68,68.84,71.49,72.18,73.16,73.31,76.27,76.8,78.62,78.7,82.03,85.44,86.05,86.28,88.25],"yaxis":"y","type":"scatter"},{"customdata":[["TruthfulQA",39.76,"bigscience\u002fbloom"],["TruthfulQA",61.56,"quantumaikr\u002fllama-2-70b-fb16-orca-chat-10k"],["TruthfulQA",65.81,"uni-tianyan\u002fUni-TianYan"],["TruthfulQA",68.26,"mistralai\u002fMistral-7B-Instruct-v0.2"],["TruthfulQA",70.49,"rishiraj\u002fmeow"],["TruthfulQA",76.73,"Sao10K\u002fSOLAR-10.7B-NahIdWin"],["TruthfulQA",78.02,"yunconglong\u002fTruthful_DPO_TomGrc_FusionNet_7Bx2_MoE_13B"],["TruthfulQA",79.36,"daxiongshu\u002fPluto_24B_DPO_63"],["TruthfulQA",79.79,"yam-peleg\u002fExperiment21-7B"],["TruthfulQA",80.88,"4season\u002falignment_model_test"]],"hovertemplate":"Model Name: %{customdata[2]}\u003cbr\u003eMetric Name: %{customdata[0]}\u003cbr\u003eDate: %{x}\u003cbr\u003eMetric Value: %{y}","legendgroup":"TruthfulQA","line":{"color":"#19d3f3","dash":"solid"},"marker":{"symbol":"circle"},"mode":"lines+markers","name":"TruthfulQA","orientation":"v","showlegend":true,"x":["2023-08-25T13:05:50+00:00","2023-09-09T10:52:17+00:00","2023-10-16T12:46:18+00:00","2023-12-11T20:10:06+00:00","2023-12-14T06:54:16+00:00","2023-12-18T13:38:24+00:00","2024-01-21T09:10:58+00:00","2024-01-25T12:27:00+00:00","2024-02-22T11:52:10+00:00","2024-03-16T11:32:40+00:00"],"xaxis":"x","y":[39.76,61.56,65.81,68.26,70.49,76.73,78.02,79.36,79.79,80.88],"yaxis":"y","type":"scatter"}], {"template":{"data":{"histogram2dcontour":[{"type":"histogram2dcontour","colorbar":{"outlinewidth":0,"ticks":""},"colorscale":[[0.0,"#0d0887"],[0.1111111111111111,"#46039f"],[0.2222222222222222,"#7201a8"],[0.3333333333333333,"#9c179e"],[0.4444444444444444,"#bd3786"],[0.5555555555555556,"#d8576b"],[0.6666666666666666,"#ed7953"],[0.7777777777777778,"#fb9f3a"],[0.8888888888888888,"#fdca26"],[1.0,"#f0f921"]]}],"choropleth":[{"type":"choropleth","colorbar":{"outlinewidth":0,"ticks":""}}],"histogram2d":[{"type":"histogram2d","colorbar":{"outlinewidth":0,"ticks":""},"colorscale":[[0.0,"#0d0887"],[0.1111111111111111,"#46039f"],[0.2222222222222222,"#7201a8"],[0.3333333333333333,"#9c179e"],[0.4444444444444444,"#bd3786"],[0.5555555555555556,"#d8576b"],[0.6666666666666666,"#ed7953"],[0.7777777777777778,"#fb9f3a"],[0.8888888888888888,"#fdca26"],[1.0,"#f0f921"]]}],"heatmap":[{"type":"heatmap","colorbar":{"outlinewidth":0,"ticks":""},"colorscale":[[0.0,"#0d0887"],[0.1111111111111111,"#46039f"],[0.2222222222222222,"#7201a8"],[0.3333333333333333,"#9c179e"],[0.4444444444444444,"#bd3786"],[0.5555555555555556,"#d8576b"],[0.6666666666666666,"#ed7953"],[0.7777777777777778,"#fb9f3a"],[0.8888888888888888,"#fdca26"],[1.0,"#f0f921"]]}],"heatmapgl":[{"type":"heatmapgl","colorbar":{"outlinewidth":0,"ticks":""},"colorscale":[[0.0,"#0d0887"],[0.1111111111111111,"#46039f"],[0.2222222222222222,"#7201a8"],[0.3333333333333333,"#9c179e"],[0.4444444444444444,"#bd3786"],[0.5555555555555556,"#d8576b"],[0.6666666666666666,"#ed7953"],[0.7777777777777778,"#fb9f3a"],[0.8888888888888888,"#fdca26"],[1.0,"#f0f921"]]}],"contourcarpet":[{"type":"contourcarpet","colorbar":{"outlinewidth":0,"ticks":""}}],"contour":[{"type":"contour","colorbar":{"outlinewidth":0,"ticks":""},"colorscale":[[0.0,"#0d0887"],[0.1111111111111111,"#46039f"],[0.2222222222222222,"#7201a8"],[0.3333333333333333,"#9c179e"],[0.4444444444444444,"#bd3786"],[0.5555555555555556,"#d8576b"],[0.6666666666666666,"#ed7953"],[0.7777777777777778,"#fb9f3a"],[0.8888888888888888,"#fdca26"],[1.0,"#f0f921"]]}],"surface":[{"type":"surface","colorbar":{"outlinewidth":0,"ticks":""},"colorscale":[[0.0,"#0d0887"],[0.1111111111111111,"#46039f"],[0.2222222222222222,"#7201a8"],[0.3333333333333333,"#9c179e"],[0.4444444444444444,"#bd3786"],[0.5555555555555556,"#d8576b"],[0.6666666666666666,"#ed7953"],[0.7777777777777778,"#fb9f3a"],[0.8888888888888888,"#fdca26"],[1.0,"#f0f921"]]}],"mesh3d":[{"type":"mesh3d","colorbar":{"outlinewidth":0,"ticks":""}}],"scatter":[{"fillpattern":{"fillmode":"overlay","size":10,"solidity":0.2},"type":"scatter"}],"parcoords":[{"type":"parcoords","line":{"colorbar":{"outlinewidth":0,"ticks":""}}}],"scatterpolargl":[{"type":"scatterpolargl","marker":{"colorbar":{"outlinewidth":0,"ticks":""}}}],"bar":[{"error_x":{"color":"#2a3f5f"},"error_y":{"color":"#2a3f5f"},"marker":{"line":{"color":"#E5ECF6","width":0.5},"pattern":{"fillmode":"overlay","size":10,"solidity":0.2}},"type":"bar"}],"scattergeo":[{"type":"scattergeo","marker":{"colorbar":{"outlinewidth":0,"ticks":""}}}],"scatterpolar":[{"type":"scatterpolar","marker":{"colorbar":{"outlinewidth":0,"ticks":""}}}],"histogram":[{"marker":{"pattern":{"fillmode":"overlay","size":10,"solidity":0.2}},"type":"histogram"}],"scattergl":[{"type":"scattergl","marker":{"colorbar":{"outlinewidth":0,"ticks":""}}}],"scatter3d":[{"type":"scatter3d","line":{"colorbar":{"outlinewidth":0,"ticks":""}},"marker":{"colorbar":{"outlinewidth":0,"ticks":""}}}],"scattermapbox":[{"type":"scattermapbox","marker":{"colorbar":{"outlinewidth":0,"ticks":""}}}],"scatterternary":[{"type":"scatterternary","marker":{"colorbar":{"outlinewidth":0,"ticks":""}}}],"scattercarpet":[{"type":"scattercarpet","marker":{"colorbar":{"outlinewidth":0,"ticks":""}}}],"carpet":[{"aaxis":{"endlinecolor":"#2a3f5f","gridcolor":"white","linecolor":"white","minorgridcolor":"white","startlinecolor":"#2a3f5f"},"baxis":{"endlinecolor":"#2a3f5f","gridcolor":"white","linecolor":"white","minorgridcolor":"white","startlinecolor":"#2a3f5f"},"type":"carpet"}],"table":[{"cells":{"fill":{"color":"#EBF0F8"},"line":{"color":"white"}},"header":{"fill":{"color":"#C8D4E3"},"line":{"color":"white"}},"type":"table"}],"barpolar":[{"marker":{"line":{"color":"#E5ECF6","width":0.5},"pattern":{"fillmode":"overlay","size":10,"solidity":0.2}},"type":"barpolar"}],"pie":[{"automargin":true,"type":"pie"}]},"layout":{"autotypenumbers":"strict","colorway":["#636efa","#EF553B","#00cc96","#ab63fa","#FFA15A","#19d3f3","#FF6692","#B6E880","#FF97FF","#FECB52"],"font":{"color":"#2a3f5f"},"hovermode":"closest","hoverlabel":{"align":"left"},"paper_bgcolor":"white","plot_bgcolor":"#E5ECF6","polar":{"bgcolor":"#E5ECF6","angularaxis":{"gridcolor":"white","linecolor":"white","ticks":""},"radialaxis":{"gridcolor":"white","linecolor":"white","ticks":""}},"ternary":{"bgcolor":"#E5ECF6","aaxis":{"gridcolor":"white","linecolor":"white","ticks":""},"baxis":{"gridcolor":"white","linecolor":"white","ticks":""},"caxis":{"gridcolor":"white","linecolor":"white","ticks":""}},"coloraxis":{"colorbar":{"outlinewidth":0,"ticks":""}},"colorscale":{"sequential":[[0.0,"#0d0887"],[0.1111111111111111,"#46039f"],[0.2222222222222222,"#7201a8"],[0.3333333333333333,"#9c179e"],[0.4444444444444444,"#bd3786"],[0.5555555555555556,"#d8576b"],[0.6666666666666666,"#ed7953"],[0.7777777777777778,"#fb9f3a"],[0.8888888888888888,"#fdca26"],[1.0,"#f0f921"]],"sequentialminus":[[0.0,"#0d0887"],[0.1111111111111111,"#46039f"],[0.2222222222222222,"#7201a8"],[0.3333333333333333,"#9c179e"],[0.4444444444444444,"#bd3786"],[0.5555555555555556,"#d8576b"],[0.6666666666666666,"#ed7953"],[0.7777777777777778,"#fb9f3a"],[0.8888888888888888,"#fdca26"],[1.0,"#f0f921"]],"diverging":[[0,"#8e0152"],[0.1,"#c51b7d"],[0.2,"#de77ae"],[0.3,"#f1b6da"],[0.4,"#fde0ef"],[0.5,"#f7f7f7"],[0.6,"#e6f5d0"],[0.7,"#b8e186"],[0.8,"#7fbc41"],[0.9,"#4d9221"],[1,"#276419"]]},"xaxis":{"gridcolor":"white","linecolor":"white","ticks":"","title":{"standoff":15},"zerolinecolor":"white","automargin":true,"zerolinewidth":2},"yaxis":{"gridcolor":"white","linecolor":"white","ticks":"","title":{"standoff":15},"zerolinecolor":"white","automargin":true,"zerolinewidth":2},"scene":{"xaxis":{"backgroundcolor":"#E5ECF6","gridcolor":"white","linecolor":"white","showbackground":true,"ticks":"","zerolinecolor":"white","gridwidth":2},"yaxis":{"backgroundcolor":"#E5ECF6","gridcolor":"white","linecolor":"white","showbackground":true,"ticks":"","zerolinecolor":"white","gridwidth":2},"zaxis":{"backgroundcolor":"#E5ECF6","gridcolor":"white","linecolor":"white","showbackground":true,"ticks":"","zerolinecolor":"white","gridwidth":2}},"shapedefaults":{"line":{"color":"#2a3f5f"}},"annotationdefaults":{"arrowcolor":"#2a3f5f","arrowhead":0,"arrowwidth":1},"geo":{"bgcolor":"white","landcolor":"#E5ECF6","subunitcolor":"white","showland":true,"showlakes":true,"lakecolor":"white"},"title":{"x":0.05},"mapbox":{"style":"light"}}},"xaxis":{"anchor":"y","domain":[0.0,1.0],"title":{"text":"date"}},"yaxis":{"anchor":"x","domain":[0.0,1.0],"title":{"text":"score"},"range":[0,100]},"legend":{"title":{"text":"task"},"tracegroupgap":0,"x":0.8,"y":0,"traceorder":"normal"},"title":{"text":"Top Scores and Human Baseline Over Time (from last update)"},"shapes":[{"line":{"color":"#636efa","dash":"dot"},"type":"line","x0":0,"x1":1,"xref":"x domain","y0":80.0,"y1":80.0,"yref":"y"},{"line":{"color":"#ab63fa","dash":"dot"},"type":"line","x0":0,"x1":1,"xref":"x domain","y0":95.0,"y1":95.0,"yref":"y"},{"line":{"color":"#EF553B","dash":"dot"},"type":"line","x0":0,"x1":1,"xref":"x domain","y0":89.8,"y1":89.8,"yref":"y"},{"line":{"color":"#19d3f3","dash":"dot"},"type":"line","x0":0,"x1":1,"xref":"x domain","y0":94.0,"y1":94.0,"yref":"y"},{"line":{"color":"#00cc96","dash":"dot"},"type":"line","x0":0,"x1":1,"xref":"x domain","y0":94.0,"y1":94.0,"yref":"y"},{"line":{"color":"#FFA15A","dash":"dot"},"type":"line","x0":0,"x1":1,"xref":"x domain","y0":100,"y1":100,"yref":"y"}],"annotations":[{"font":{"color":"#636efa","size":10},"showarrow":false,"text":"ARC human baseline","x":0,"xanchor":"left","xref":"x domain","y":80.0,"yanchor":"top","yref":"y"},{"font":{"color":"#ab63fa","size":10},"showarrow":false,"text":"HellaSwag human baseline","x":0,"xanchor":"left","xref":"x domain","y":95.0,"yanchor":"bottom","yref":"y"},{"font":{"color":"#EF553B","size":10},"showarrow":false,"text":"MMLU human baseline","x":0,"xanchor":"left","xref":"x domain","y":89.8,"yanchor":"top","yref":"y"},{"font":{"color":"#19d3f3","size":10},"showarrow":false,"text":"TruthfulQA human baseline","x":0,"xanchor":"left","xref":"x domain","y":94.0,"yanchor":"top","yref":"y"},{"font":{"color":"#00cc96","size":10},"showarrow":false,"text":"Winogrande human baseline","x":0,"xanchor":"left","xref":"x domain","y":94.0,"yanchor":"top","yref":"y"},{"font":{"color":"#FFA15A","size":10},"showarrow":false,"text":"GSM8K human baseline","x":0,"xanchor":"left","xref":"x domain","y":100,"yanchor":"top","yref":"y"}]}, {"responsive": true} ) }; </script> </div> |