Text Classification
Transformers
Safetensors
English
nli
cross-encoder
qwen3.5
reranker
image-text-to-text
Instructions to use AlexWortega/openjev with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use AlexWortega/openjev with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-classification", model="AlexWortega/openjev")# pip install -U transformers accelerate # Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("AlexWortega/openjev", device_map="auto") - Notebooks
- Google Colab
- Kaggle
Commit History
-
Add FinCalc-NLI eval (financial arithmetic) + results for all checkpoints
Verified signatureHi Alex, thanks for openjev. I use it as the first stage of a judge cascade for banking agents, and it's excellent on lookups and faithfulness. I found one gap: claims that need arithmetic (ratios, covenant tests, net figures). This PR adds `code/eval_fincalc.py`, in the style of `eval_extra.py` and built on your `NLIScorer`, unchanged. It evaluates any checkpoint on [FinCalc-NLI](https://huggingface.co/datasets/anespo28/fincalc-nli) (MIT, synthetic, labels computed in code). It also adds one line to "What's inside". Results on the 4,000-item test split (run on HF Jobs with your `code/eval.py` as-is): | checkpoint | FinCalc acc | arithmetic slips caught | wrong-entity figures caught | covenant pass/breach | |---|---|---|---|---| | 4B v5 | 63.2% | 30% | 33% | 65% | | 4B v2 | 58.2% | 34% | 39% | 17% | | 0.8B v2s-long | 50.5% | 12% | 7% | 32% | | 4B original | 31.9% | 0% | 0% | 0% | A LoRA on the original 4B with general-NLI replay reaches 94.4% ([openjev-fincalc-4b](https://huggingfa…b639183
-
add qwen3.5-4b-nli-v5-nvfp4 (ModelOpt NVFP4 PTQ + QAD) with SGLang bench
Verified signature058a6c2
-
card: links to the 4B / 2B / 0.8B v5 checkpoints
Verified signature552759d -
add qwen3.5-0.8b-nli-v5
Verified signaturec64009c -
add qwen3.5-2b-nli-v5
Verified signaturedfdc1cb -
card: openjev-4B v5 (JevBench 0.814 public, judging/faithfulness/IF tables, determinism, injection limits, contamination manifest)
Verified signaturee9b1201 -
add contamination manifest (train+test splits used for v5 panel)
Verified signature7ac2931 -
add RESULTS-v5.md
Verified signaturec59abe0 -
code: gen_ifeval.py
Verified signaturee368d07 -
code: think_probe.py
Verified signature3eced74 -
code: eval.py
Verified signatureb35f155 -
code: train.py
Verified signatureb6aa6e6 -
code: distill_teacher.py
Verified signature427c2b8 -
code: hard_gen.py
Verified signaturea9bf529 -
code: data_mix.py
Verified signatureb5bd9a2 -
code: webql_null.py
Verified signature1ccf050 -
code: eval_agentic.py
Verified signaturef2e1333 -
code: eval_extra.py
Verified signature3909a82 -
code: eval_security.py
Verified signatureb390a79 -
code: order_test.py
Verified signature8e01204 -
code: eval_jevbench.py
Verified signaturebe3484d -
code: modeling_openjev.py
Verified signature106dab5 -
code: openjev_decide.py
Verified signature4d701a9 -
4B v5: typed-decision cross-encoder (panel + gpt-5.5 hard items + teacher distillation)
Verified signature5df9876
-
SGLang serving: external model package, launcher, client, benchmark; card section
Verified signaturef004f37
-
Document the 0.8B v2s-long checkpoint served by the Space
Verified signature4395b29 -
upload openjev-0.8B v2s-long (stage3, 4k ctx, faith/ifollow/false-premise + long-doc)
Verified signature0271e1a -
Share Qwen3.5 prefix cache across NLI hypotheses (#1)
- Share Qwen3.5 prefix cache across NLI hypotheses (e93f6db8b166525b8d4cf7778af6e503b0dea49d) - Use pairwise batch for one or two hypotheses (c959f4d3bbc87cc22403f9028f430490f42dabc5) Co-authored-by: David Schiff <epsilon3@users.noreply.huggingface.co>65ae5e7
-
upload qwen3.5-35b-a3b-nli
Verified signature4b5f9a6 -
upload qwen3.5-4b-nli-v2
Verified signature9f176ba -
openjev v2: results, videos (Doom v2, Minecraft), code, card
Verified signature0988d66
-
add results/webql_geminiprompt_4b.json
Verified signaturef8187e6 -
add results/webql_bench_4b.json
Verified signature466a519 -
add results/webql_scaffold_4b.json
Verified signature7c326a9 -
add results/webql_mlp_4b.json
Verified signature9d9aabe -
add results/webql_fulldoc_35b.json
Verified signature139fb28 -
add results/webql_mlp_35b.json
Verified signaturec113a3e -
add results/webql_bench_35b.json
Verified signature90d97ca -
add results/webql_fulldoc_4b.json
Verified signature06d1cbe -
add results/webql_geminiprompt_35b.json
Verified signature4473830 -
add results/latent_mlp_35b.json
Verified signature3d5cbef -
add results/qwen35b_all.json
Verified signature9cba221 -
add assets/radar_openjev_35b.png
Verified signatured785095 -
add code/webql_gemini_prompt.py
Verified signature5a0cd95 -
add code/webql_fulldoc.py
Verified signatureed99575 -
add code/webql_mlp.py
Verified signature7f48358 -
add code/webql_scaffold.py
Verified signature6053aa9 -
add code/webql_bench.py
Verified signature8ce2faf -
add code/latent_mlp.py
Verified signature0027527 -
add code/radar35.py
Verified signature18c48ba