Open Problems

Question Answering & Reading Comprehension

Multilingual Robustness of Table Question Answering Models

Scope to testPartly addressed
Possible candidate · 2/5 runs3 papers report this67% from 2025+

Generated automatically from the limitations stated in 3 papers (ACL, EMNLP, NAACL), listed under Evidence. It is not a paper, and it does not come from papers submitted to CSPaper.

The problem

Current table question answering benchmarks and systems predominantly evaluate questions, tables, and associated text in a single language—typically English or Simplified Chinese. As a result, it is unknown how robust existing table QA architectures are when queries, table contents, or schema descriptions span multiple languages or differ linguistically from one another. Without cross-lingual and multilingual evaluation, practitioners cannot determine whether standard table parsing, semantic parsing, and retrieval-augmented methods generalize to non-English structured data or fail due to language mismatch.

Why it matters

It provides the first systematic assessment of cross-lingual transfer and robustness in table QA, identifying whether multilingual language models natively understand non-English structured representations or require specialized cross-lingual table pre-training.

Ways to approach it

Prior-work checks are free with an account. Results someone already ran are shown to everyone.

  1. 1

    Translate and culturally localize existing English table QA benchmarks (e.g., WikiTableQuestions, FeTaQA) into 6–10 typologically diverse languages using professional translation with human quality checks, then measure the zero-shot and fine-tuned performance degradation of standard table QA models.

  2. 2

    Construct a mixed-language table QA evaluation suite using inherently multilingual tables (e.g., from Wikidata or multilingual Wikipedia) paired with native-language questions across different scripts, measuring accuracy across aligned vs. mismatched query-table language pairs.

Have a different approach?

Describe how you would tackle this problem and we'll look for papers that already do it. Free; your text stays private.

Free · 3 checks per day

Why it might fail

Machine-translating table headers and contents could introduce cell alignment artifacts that obscure model QA capabilities rather than measuring true multilingual comprehension, or proprietary multilingual LLMs might already exhibit near-lossless zero-shot cross-lingual table reasoning on standard languages.

Evidence

Each paper's own statement of the limitation, verbatim.

Nearest existing work

Related open problems

Question Answering & Reading Comprehension

Effect to explainOpen

Principled Allocation of Low-Rank Adaptation Capacity When Full Fine-Tuning Is Unaffordable

LoRA is the default substitute for full fine-tuning, yet across tasks it produces large, sometimes catastrophic, performance drops relative to full updates — and practitioners currently have no way to predict or prevent them. The failures stem from arbitrary capacity choices: uniform ranks across layers and clients, tiny adapter budgets, and no guidance on when the low-rank constraint itself (rather than the task) is responsible. Until rank and layer allocation are treated as first-class decisions, LoRA deployments remain gambles that either waste compute or silently degrade quality.

Strong candidate · 4/5 runs6 papers report this33% from 2025+
Generated automatically, not curated by hand. Automated prior-work checks catch about a third of existing work, so treat this problem as a lead to investigate.