Science Question Answering on ARC Challenging (Best accuracy, Average accuracy)
96.3Best AccuracyCISC
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CISCLLM Backbone=Llama3.3 70B2026.05 | 96.3 | 96.3 | |
| VecCISC + KMeansLLM Backbone=Llama3.3 70B2026.05 | 96.3 | 96.3 | |
| VecCISC + HACLLM Backbone=Llama3.3 70B2026.05 | 96.3 | 96.3 | |
| VecCISC + KMeansLLM Backbone=GPT 4o-mini2026.05 | 96.1 | 96 | |
| VecCISC + HACLLM Backbone=GPT 4o-mini2026.05 | 96.1 | 96.1 | |
| CISCLLM Backbone=GPT 4o-mini2026.05 | 96 | 96 | |
| VecCISC (random)LLM Backbone=Llama3.3 70B2026.05 | 96 | 95.6 | |
| SC BaselineLLM Backbone=Llama3.3 70B2026.05 | 95.9 | 95.9 | |
| SC BaselineLLM Backbone=GPT 4o-mini2026.05 | 95.7 | 95.7 | |
| VecCISC (random)LLM Backbone=GPT 4o-mini2026.05 | 95.6 | 95.2 | |
| CISCLLM Backbone=Qwen2.5 7B2026.05 | 93.2 | 93.2 | |
| VecCISC + KMeansLLM Backbone=Qwen2.5 7B2026.05 | 93.2 | 93.2 | |
| VecCISC + HACLLM Backbone=Qwen2.5 7B2026.05 | 93.2 | 93.2 | |
| SC BaselineLLM Backbone=Qwen2.5 7B2026.05 | 92.9 | 92.9 | |
| CISCLLM Backbone=Llama3.1 8B2026.05 | 91.1 | 91.1 | |
| VecCISC + KMeansLLM Backbone=Llama3.1 8B2026.05 | 91.1 | 91.1 | |
| VecCISC + HACLLM Backbone=Llama3.1 8B2026.05 | 91.1 | 91.1 | |
| SC BaselineLLM Backbone=Llama3.1 8B2026.05 | 91 | 91 | |
| VecCISC (random)LLM Backbone=Qwen2.5 7B2026.05 | 90.9 | 90.2 | |
| VecCISC (random)LLM Backbone=Llama3.1 8B2026.05 | 75.5 | 74.8 | |
| SC BaselineLLM Backbone=Mistral 7B2026.05 | 71.3 | 71.3 | |
| CISCLLM Backbone=Mistral 7B2026.05 | 71.3 | 71.3 | |
| VecCISC + KMeansLLM Backbone=Mistral 7B2026.05 | 71.3 | 71.3 | |
| VecCISC + HACLLM Backbone=Mistral 7B2026.05 | 71.3 | 71.3 | |
| VecCISC (random)LLM Backbone=Mistral 7B2026.05 | 54.9 | 52.5 |