Skip to content

Benchmark

BABILong qa1 0k

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

91%
Mean
99%
Median
14.4%
Std dev
60%
10th percentile
100%
90th percentile
54%
Min
54%100%
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-tuneqa1, 0k· BABILong v0 data docs100%
~ Mamba (130M) fine-tuneqa1, 0k· BABILong v0 data docs100%
~ RMT (137M) fine-tuneqa1, 0k· BABILong v0 data docs100%
ChatGLM3 6Bchatglm3-6b-128k, qa1, 0k, 128k· BABILong v0 data docs82%
GPT-4-0125-previewqa1, 0k· BABILong v0 data docs99%
Llama 2 7bLLaMA-2-7B-32K, qa1, 0k, 32k· BABILong v0 data docs54%
Llama 3 8B InstructMeta-Llama-3-8B-Instruct, qa1, 0k· BABILong v0 data docs98%
Llama3-ChatQA-1.5-8B + RAGqa1, 0k· BABILong v0 data docs60%
Mistral 7B Instruct v0.2Mistral-7b-Instruct-v0.2, qa1, 0k· BABILong v0 data docs92%
Mixtral 8x22B Instruct v0.1Mixtral-8x22B-Instruct-v0.1, qa1, 0k· BABILong v0 data docs100%
Mixtral 8x7B Instruct v0.1Mixtral-8x7B-Instruct-v0.1, qa1, 0k· BABILong v0 data docs99%
Phi-3 Medium 128K InstructPhi-3-medium-128k-instruct, qa1, 0k· BABILong v0 data docs100%
Phi-3 Mini 128K InstructPhi-3-mini-128k-instruct, qa1, 0k· BABILong v0 data docs97%
Yi-34B01-ai/Yi-34B-200k, qa1, 0k, 200k· BABILong v0 data docs85%
Yi-34B01-ai/Yi-34B, qa1, 0k· BABILong v0 data docs99%

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