Eklav-Table-8B

Qwen3-8B for single-call listwise table reranking: given a question and candidate tables, it returns the full ranking in one generation. Trained with Eklav: the teacher's reasoning is given as context during training and the loss is on the final ranking only; at inference it writes its own, shorter reasoning. Code · Std-SFT · Naive-SFT

Results

	SQA	TAT-QA	TabFact	NQ-Tables	AITQA	FeTaQA	OTT-QA	OpenWikiTables
Qwen3-8B (base)	69.51	50.51	60.03	72.10	77.04	89.63	81.27	88.68
Naive-SFT	71.41	49.59	62.20	68.88	68.23	91.42	88.25	91.98
Std-SFT	72.05	52.03	66.61	74.73	88.25	88.06	83.20	90.29
▶ Eklav	73.81	54.59	68.90	72.20	85.99	91.95	90.27	93.78

Input / output format

Input — a chat message with the question followed by each candidate table, labeled ### Table 1, ### Table 2, ...:

Question: Which table shows 2022 quarterly revenue by region?

### Table 1
| Region | Q1 2022 | Q2 2022 | Q3 2022 | Q4 2022 |
|---|---|---|---|---|
| North America | 120 | 134 | 128 | 145 |
| Europe | 88 | 91 | 95 | 102 |

### Table 2
| Product | Units Sold | Year |
|---|---|---|
| Widget A | 4200 | 2021 |

### Table 3
| Region | Headcount |
|---|---|
| North America | 340 |

Output — a <think> block with the model's reasoning, followed by a single JSON object with the ranked, one-indexed candidate positions, best first:

<think>
Table 1 has quarterly revenue by region for 2022, which is exactly what the question asks
for. Table 3 has region data but no revenue. Table 2 has neither region nor 2022 data.
</think>
{"ranked_tables": [1, 3, 2]}

Usage with vLLM

from vllm import LLM, SamplingParams
from transformers import AutoTokenizer

repo = "AdarshSingh7647/Eklav-Table-8B"
tok = AutoTokenizer.from_pretrained(repo)
llm = LLM(model=repo, dtype="bfloat16", max_model_len=32768)

system = ("You are a table relevance expert. Given a question and a set of candidate tables "
          "rank them from most to least useful for answering the question. Reason in a "
          "<think>...</think> block then output exactly JSON with key ranked_tables.")
user = "Question: ...\n\n### Table 1\n...\n\n### Table 2\n...\n"

msgs = [{"role": "system", "content": system}, {"role": "user", "content": user}]
text = tok.apply_chat_template(msgs, tokenize=False, add_generation_prompt=True)
out = llm.generate([text], SamplingParams(temperature=0.6, top_p=0.95, max_tokens=8192))
print(out[0].outputs[0].text)

Citation

@misc{singh2026tabrank,
      title={TabRank: Chain-of-Thought Distillation for Table Re-Rankers},
      author={Adarsh Singh and Kushal Raj Bhandari and Jianxi Gao and Soham Dan and Vivek Gupta},
      year={2026},
      eprint={2607.25182},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2607.25182}
}