Text Generation
Text Generation is the task of generating text with the goal of appearing indistinguishable to human-written text. This task is more formally known as "natural language generation" in the literature.
Text generation can be addressed with Markov processes or deep generative models like LSTMs. Recently, some of the most advanced methods for text generation include BART, GPT and other GAN-based approaches. Text generation systems are evaluated either through human ratings or automatic evaluation metrics like METEOR, ROUGE, and BLEU.
Further readings:
( Image credit: Adversarial Ranking for Language Generation )
Papers
Showing 1–10 of 5335 papers
All datasetsDARTCOCO CaptionsEMNLP2017 WMTReDialCommonGenROCStoriesChinese PoemsCzech restaurant informationOpenWebTextSciQYahoo QuestionsADGEN
Benchmark Results
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | T5B Baseline | BLEU | 48.74 | — | Unverified |
| 2 | FactT5B | BLEU | 48.37 | — | Unverified |
| 3 | JointGT Baseline | BLEU | 47.51 | — | Unverified |
| 4 | FactJointGT | BLEU | 47.39 | — | Unverified |
| 5 | Control Prefixes (T5-large) | METEOR | 0.41 | — | Unverified |
| 6 | T5 | METEOR | 0.12 | — | Unverified |
| 7 | BART | METEOR | 0.11 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | UniLM | CIDEr | 14.92 | — | Unverified |
| 2 | BART (TextBox 2.0) | CIDEr | 12.98 | — | Unverified |
| 3 | BART | METEOR | 0.3 | — | Unverified |
| 4 | T5 | METEOR | 0.29 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | Beam search + A*esque (beam) | BLEU-1 | 34.4 | — | Unverified |
| 2 | Beam search + A*esque (sample) | BLEU-1 | 34.4 | — | Unverified |
| 3 | Beam search + A*esque (greedy) | BLEU-1 | 34.3 | — | Unverified |
| 4 | Beam search | BLEU-1 | 33.7 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | GPT2-124M | eval_loss | 3.12 | — | Unverified |
| 2 | GPT2-81M-LOOP | eval_loss | 3.11 | — | Unverified |
| 3 | GPT2-Hermite | eval_loss | 2.91 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | LLaMA-65B+CFG (zero-shot) | Accuracy | 96.6 | — | Unverified |
| 2 | LLaMA-30B+CFG (zero-shot) | Accuracy | 96.4 | — | Unverified |
| 3 | LLaMA-13B+CFG (zero-shot) | Accuracy | 95.1 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | CNN-VAE | NLL | 332.1 | — | Unverified |
| 2 | SA-VAE | NLL | 327.5 | — | Unverified |
| 3 | Aggressive VAE | NLL | 326.7 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | BART (TextBox 2.0) | BLEU-4 | 10.2 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | STWGAN-GP | BLEU-3 | 0.62 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | PALM | ROUGE-L | 41.41 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | BART (TextBox 2.0) | ROUGE-L | 64.34 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | AEM+Attention | BLEU-1 | 14.17 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | GPT-4 | ASR | 65.1 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | BART (TextBox 2.0) | ROUGE-L | 42.96 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | Graph2Seq | BLEU | 22 | — | Unverified |
| # | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| 1 | WGANGP + DGflow | JS-4 | 0.19 | — | Unverified |