← All work

LLM research · reproduction · domain adaptation

TISER Temporal Reasoning

A reproduction of ACL 2025 TISER on Qwen2.5-7B, extended with a context-memory conflict probe and a professional-tennis temporal reasoning adaptation study.

Year
2026
Role
Collaborative research; contributed tennis-domain adaptation and integration of the main / tennis training-evaluation workflows

Collaboration note — Collaborative project. Repository hosted by a collaborator; ownership is not presented as solo work.

TISER pipeline showing structured temporal traces, LoRA fine-tuning, baseline evaluation, context-memory conflict analysis and tennis domain adaptation.
Diagram based on the repository architecture and reported evaluation path — not a fabricated product screenshot.

Problem

The project asks two questions beyond simply running an LLM benchmark: can the structured temporal-reasoning target from TISER be faithfully reproduced, and does that reasoning behaviour transfer when the model is pushed into conflicting context or a new domain?

What I built

The repository fine-tunes Qwen2.5-7B-Instruct with LoRA on the released TISER training data, evaluates exact match and token F1 on the full benchmark, then adds two experimental branches: a context-memory conflict probe and a tennis adaptation pipeline with audited synthetic examples and continued fine-tuning.

Full-test reproduction0.949 F1

Macro F1 across the five in-domain splits; macro EM 0.878 over 22,014 test examples.

Engineering decisions

Where the project becomes more than a demo.

The details that matter are usually in the constraints: who is allowed to decide, what gets measured, what is kept separate, and how failure is represented.

01

Reproduce before extending.

The baseline uses the released prompt/output format and full training/evaluation path before interpreting extension results.

02

Keep evaluation split-aware.

Five in-domain datasets define the reported macro; the OOD ToT-semantic split is reported separately rather than blended into a flattering aggregate.

03

Test failure modes, not just scores.

The context-memory study measures whether the model follows edited evidence over memorized facts and audits whether the reflection actually identifies the contradiction.

04

Treat domain adaptation as an experiment.

Tennis examples are audited, converted into the same structured format and compared across base, transferred and continued-adaptation conditions.

Evidence

What the repository actually supports.

0.878macro EM

five in-domain splits on the full TISER test set

0.949macro F1

22,014 total evaluated examples

0.732tennis EM

7B TISER continued adaptation; F1 0.856

4.2%reflection catch rate

rows where the reflection explicitly names the context-memory contradiction

Limits

What I am not claiming.

  • The baseline reproduction does not exactly match every number in the paper; the repository reports the reproduced numbers rather than hiding the gap.
  • The OOD ToT-semantic split is substantially harder and is excluded from the in-domain macro by design.
  • The conflict study shows that a structured reflection can improve behaviour without reliably verbalising the contradiction.

Signal

What this project demonstrates.

LoRA / PEFTQwen modelsresearch reproductionLLM evaluationsynthetic data pipelinesdomain adaptation

Want to inspect the implementation?

Open repository ↗Ask me about it