Skip to content

Benchmark

BABILong qa5 8k

Long context benchmarks measure how well models use information in long inputs. Scores are percentages. Higher is better.

Models and systems measured
14
Scores recorded
15
Sources
1
IQ weight
No weight published

Score distribution

76.5%
Mean
73%
Median
14.5%
Std dev
59%
10th percentile
99%
90th percentile
53%
Min
53%99%
The band covers the middle 80% of scores. The line marks the median.

Results

The source marks these results as descriptive, not capability scores, so we list them by name instead of ranking them.

Model or systemScore
~ ARMT (137M) fine-tuneqa5, 8k· BABILong v0 data docs98.2%
~ Mamba (130M) fine-tuneqa5, 8k· BABILong v0 data docs99%
~ RMT (137M) fine-tuneqa5, 8k· BABILong v0 data docs99%
ChatGLM3 6Bchatglm3-6b-128k, qa5, 8k, 128k· BABILong v0 data docs61%
GPT-4-0125-previewqa5, 8k· BABILong v0 data docs93%
Llama 2 7bLLaMA-2-7B-32K, qa5, 8k, 32k· BABILong v0 data docs59%
Llama 3 8B InstructMeta-Llama-3-8B-Instruct, qa5, 8k· BABILong v0 data docs73%
Llama3-ChatQA-1.5-8B + RAGqa5, 8k· BABILong v0 data docs78%
Mistral 7B Instruct v0.2Mistral-7b-Instruct-v0.2, qa5, 8k· BABILong v0 data docs69%
Mixtral 8x22B Instruct v0.1Mixtral-8x22B-Instruct-v0.1, qa5, 8k· BABILong v0 data docs69%
Mixtral 8x7B Instruct v0.1Mixtral-8x7B-Instruct-v0.1, qa5, 8k· BABILong v0 data docs66%
Phi-3 Medium 128K InstructPhi-3-medium-128k-instruct, qa5, 8k· BABILong v0 data docs82%
Phi-3 Mini 128K InstructPhi-3-mini-128k-instruct, qa5, 8k· BABILong v0 data docs72%
Yi-34B01-ai/Yi-34B-200k, qa5, 8k, 200k· BABILong v0 data docs76%
Yi-34B01-ai/Yi-34B, qa5, 8k· BABILong v0 data docs53%

Data credits: RMT-team, Kuratov et al. Official v0 numeric CSV; results licence not separately stated.