AvoCahDoe
/

qwen2-vl-7b-rlmpq-balanced

vision-language-model

Model card Files Files and versions

RL-MPQ Balanced — Qwen/Qwen2-VL-7B-Instruct

RL-MPQ fake-quantized VLM evaluation artifacts.


Repo	`AvoCahDoe/qwen2-vl-7b-rlmpq-balanced`
Scenario	Balanced
Avg bits	3.39
Collection	RL-MPQ VLM — Qwen2-VL-7B
Full results	dataset

Benchmark table (group comparison)

Model	avg_bits	group	MMMU	MMBench	ScienceQA	Avg	ΔMMMU	ΔMMBench	ΔScienceQA	ΔAvg
FP16 (baseline)	16.0	qwen2vl_balanced	49.67	78.1	85.08	70.95	0.0	0.0	0.0	0.0
INT4 (bnb NF4)	4.0	qwen2vl_balanced	39.33	21.13	75.31	45.26	-10.34	-56.97	-9.77	-25.69
RL-MPQ Balanced	3.39	qwen2vl_balanced	46.67	75.0	84.23	68.63	-3.0	-3.1	-0.85	-2.32

Artifacts in this repo

artifacts/figures/ — plots for the model group
artifacts/benchmark_table.csv — FP16 / INT4 / RL-MPQ accuracies
eval_results.json — structured eval metadata (if present)

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for AvoCahDoe/qwen2-vl-7b-rlmpq-balanced

Base model

Qwen/Qwen2-VL-7B

Finetuned

Qwen/Qwen2-VL-7B-Instruct

Finetuned

(605)

this model

Collections including AvoCahDoe/qwen2-vl-7b-rlmpq-balanced

RL-MPQ LLaVA-1.5-13B VLM Evaluation

RL-MPQ fake-quantized LLaVA-1.5-13B vision-language models and their VLMEvalKit benchmark results (MMMU, MMBench, ScienceQA). • 7 items • Updated Jun 13

RL-MPQ VLM — Qwen2-VL-7B

RL-MPQ fake-quant Qwen2-VL-7B (qwen2_5_7b Balanced policy). 3.39 bpw. VLMEvalKit exact-match eval. • 2 items • Updated Jun 13